TRI-Ada'95 e-mail List
Caveat lector: this is simply a copy of all messages sent to the list.
No filter has been applied to select messages, so some very boring
messages keep the signal/noise ratio low (i.e. it's not all about
TRI-Ada'95). -- Magnus Kempe
From: brashepw@ss2.sews.wpafb.af.mil (PHILIP W. BRASHEAR)
Subject: My ego and Tri-Ada '95
Date: Fri, 3 Nov 95 08:09:01 EST
All,
This may not be significant to most people, but some were expecting to see
me at Tri-Ada. For personal reasons, I'm not able to attend. Nevertheless,
the Wednesday evening PIWG meeting will proceed, with the ACES update being
given by Tom Mittelkamp. I encourage everyone concerned with performance
testing of Ada 95 implementations to attend and to provide suggestions for
extending Ada 95-specific testing. Tom will also present my position on a
Wednesday morning panel in the Government track: Will Ada 95 Experiences
Parallel Ada 83 Experiences? My position is directed toward my areas of
"expertise": validation and performance testing (ACVC and ACES).
I hope that attendees will use this mailing list (triada95@ocsystems.com)
as it's intended: to inform people like me who can't attend.
Happy Disneylanding!
Phil Brashear
PIWG Chair
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: AdaSAGE
Date: 07 Nov 95 01:06:58 EST
The AdaSAGE meeting had more than thirty people in attendance. I joined the
meeting late. Progress is being made on linking to Windows95.
The discussion of interest to me was about the commercialization of AdaSAGE. I
do not think there is yet an adequate model for commercialization; who will pay,
who will benefit, who are the customers, who are the suppliers, what incentives
are there for investment, what will be the continuing role of INEL, and so
forth. A subcommittee was appointed (and I am a member) to come up with a policy
/ procedure for commercializing AdaSAGE. Anyone interested is welcome to contact
me.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Reuse Working Group
Date: 07 Nov 95 01:07:03 EST
Harry Joiner chaired the Reuse Working Group session Monday evening. There were
about twenty people there. They have been working with industry groups and
government policy officials involved with DoD's reuse initiative. They have an
e-mail redistribution list described in the FAQ for the PAL.
To be added to the reusewg@wunet.wustl.edu, send an untitled message saying
"subscribe reusewg" to listserv@wunet.wustl.edu.
I had gone to the meeting to discuss the possibility of developing a collection
of Ada routines that would be useful to distribute with compilers; a collection
that would be more user friendly than the extensive PAL collection. We didn't
get to that topic before my patience with the discussion of organizational
issues ran out.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Why don't people use Ada more
Date: 07 Nov 95 01:07:10 EST
Ron Oliver (sroliver@CalPoly.edu) a professor at Cal. Poly, has a theory that
people need to learn how to think at different levels. If they only handle the
low-level details of working with computers then C appeals to them. They have to
move to more abstract thinking (and this usually doesn't happen until later in
their college career) to appreciate the capabilities and facilities of Ada. It's
not as separated as this makes it sound, but basically there are a lot of things
to learn before people can work at these different levels.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Outside the sessions
Date: 07 Nov 95 01:07:13 EST
The weather in southern California was comfortable to sit outside. Of couse the
discussions turned to computations. Bevin Brett (bevin@central.co.nz) announced
that the 100 billionth hexadecimal digit of Pi is 9. He said he'd found this
while Web surfing. Discussion also turned to where we could find an appropriate
place to discuss the waterfall methodology.
This is just to remind everybody that a lot of the "fun" of attending the
conference happens outside the formal sessions.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: ASIS Information
Date: 07 Nov 95 01:07:18 EST
The first public draft of ASIS Version 2.0.E for Ada95 was provided at the ASIS
BoF on Monday evening on a floppy disk. This draft addresses the Ada95
capabilities including its new objected oriented features. Comments were
solicited on the new specification, especially on support for OO. Also on the
disk distributed were the ASIS FAQ, ASIS tutorials (for ASIS83), and a paper on
the ASIS implementation for the Ada95 GNAT compiler. ASIS users were available
to answer questions from the conference attendees. The latest ASIS information
will soon be available via the ASIS Home Page at
http://www.acm.org/sigada/WG/asiswg/asiswg.html (by 15 November).
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: ARA Poster
Date: 07 Nov 95 01:07:23 EST
The Ada Resource Association is developing a new poster, Ferruccio's "The Vision
of Ada." In a few weeks, the image will be available for viewing on the various
Ada Web pages. At TRI-Ada'95, attendees could request a copy of the poster by
filling out a card. Readers of these notes can request one by sending e-mail
with the subject "ARA Poster" to 73313.2671@compuserve.com (that's Bob Mathis
the Executive Director of the ARA). Please send the usual mailing information
plus phone, fax, and other relevant contact information.
Ferruccio Sardella, one of North America's emerging conceptual illustrators,
creates contemporary images representing interaction between people, technology
and environments. His unique approach to imagery is the result of a foundation
of study in Canada and Europe and experience helping businesses and corporations
visualize their ideas.
Ferrucio's work has been previously commissioned by Businessweek, CompuServe,
McCann Erickson, The New York Times, Psychology Today, Toronto Dominion Bank,
W.H. Freeman, and many other leading edge clients. His fame is significant and
growing as his vision and talents are appreciated all over the world.
As a separate offer, not available at the Conference, the ARA also has some
remaining copies of last year's poster showing a surfer. You can also request
one of these via e-mail. There were some requests last year that were never
filled by a company no longer working for the ARA. We have recovered their stock
and want the opportunity to try again.
Subject: Initial Message
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Date: 07 Nov 95 01:07:16 EST
This is just to remind everyone reading this list that these are my informal
comments. There are a lot of short notes so that people can respond to them as
appropriate. I have frequently missed something that I hope readers can supply.
While at the conference, I have tried to fill in things and have others help me
with corrections, but there are still things that should be amplified.
This is an experiment. We hope to learn things. Some of that learning will be
from mistakes. -- Bob Mathis
From: Angel Alvarez
Subject: Re: Initial Message
Date: Tue, 7 Nov 1995 08:51:13 +0100
Cheers from Madrid, Bob, and thank you for your wonderful
job filling us on what's going on ... I wish I was there!!!
Give my regards to Dave Emery, please.
Salud,
Angel
From: brashepw@ss2.sews.wpafb.af.mil (PHILIP W. BRASHEAR)
Subject: Re: AdaSAGE
Date: Tue, 7 Nov 95 08:05:12 EST
Bob, et. al.
It seems to me that a commercialization model for AdaSAGE would have far
more general applicability. I think immediately of the Ada Compiler Evaluation
System (ACES). Who will own it after the AJPO goes away? (Though I don't think
there's any monetary return for ACES, it still ought to be maintained and
distributed.)
Phil Brashear
PIWG Chair
From: brashepw@ss2.sews.wpafb.af.mil (PHILIP W. BRASHEAR)
Subject: Re: Why don't people use Ada more
Date: Tue, 7 Nov 95 08:13:20 EST
Ron had an Ada Letters a while back, titled something like "Pyramids and I
Igloos". He made the point (rather well) that C is fine for building igloos
(small systems that arent' expected to have long lives), but Ada is better
for building pyramids (engineering efforts that are expected to live forever).
He's got (in my opinion) an excellent grasp of such concepts, and is quite
good at expressing them. I wish more people would listen to him.
Phil Brashear
From: "Paul Whittington"
Subject: FREE: AdaSAGE CD-ROM Nov. 95 Edition
Organization: INEL Special Applications Unit
Date: Tue, 7 Nov 1995 11:08:24 MDT
If your at TriAda 95 just stop by the INEL AdaSAGE booth for a free
copy of the Nov. 95 AdaSAGE CD-ROM. Contents include:
- AdaSAGE 5.0 with WIN32 support.
- Source distributions of AdaSAGE 4.2 for Linux/GNAT
and Windows NT/GNAT.
- Hypertext, WordPerfect and ASCII documentation versions.
In the booth you can see live demos and find out what AdaSAGE is and
how it can help you deliver Ada applictions for DOS, Windows 3.1,
Windows 95, Windows NT and UNIX systems.
Come on by and say HI to Howard, Jon and Dave. Tell them Paul sent
you.
From: beidler@cs.uofs.edu (Jack Beidler)
Subject: What's Happenning
Date: Tue, 7 Nov 95 12:54:25 EST
Unfortunatley, I was unable to attend Tri-Ada do to some health
problems. Really, how is this conference going? What is the
official attendance (so far)?
Jack Beidler
From: J.M.KAMRAD.II@cdev.com (j.m.kamrad.ii)
Subject: Re: FREE: AdaSAGE CD-ROM Nov. 95 Edition
Date: Tue, 7 Nov 1995 13:01:20 -0500
>If your at TriAda 95 just stop by the INEL AdaSAGE booth for a free
>copy of the Nov. 95 AdaSAGE CD-ROM. Contents include:
>
> - AdaSAGE 5.0 with WIN32 support.
> - Source distributions of AdaSAGE 4.2 for Linux/GNAT
> and Windows NT/GNAT.
> - Hypertext, WordPerfect and ASCII documentation versions.
>
>In the booth you can see live demos and find out what AdaSAGE is and
>how it can help you deliver Ada applictions for DOS, Windows 3.1,
>Windows 95, Windows NT and UNIX systems.
>
>Come on by and say HI to Howard, Jon and Dave. Tell them Paul sent
>you.
Paul,
This sounds great but I won't be attending Tri-Ada. How might I get a copy?
Mike
------------------------------------
Mike Kamrad
Computing Devices International kamrad@cdev.com
M/S BLC W2J 1.612.921.6908
8800 Queen Avenue South FAX: 1.612.921.6552
Bloomington MN 55431
From: "Paul Whittington"
Subject: AdaSAGE CD-ROM for those not at TriAda 95
Organization: INEL Special Applications Unit
Date: Tue, 7 Nov 1995 12:03:06 MDT
For those of you who can't attend TriAda 95, like me, but still want
a new AdaSAGE CD you have two options:
1) Download any or all of the CD rom from
ftp://sageftp.inel.gov/pub/sage/cdrom002
2) E-Mail me with your land address and "IF" we have the budget
and enough CDs left we will mail them to you. I will let you
know ASAP if we can indeed mail them to you.
From: "Paul Whittington"
Subject: Re: FREE: AdaSAGE CD-ROM Nov. 95 Edition
Organization: INEL Special Applications Unit
Date: Tue, 7 Nov 1995 12:05:30 MDT
> Date sent: Tue, 7 Nov 1995 10:03:08 -0800
> Send reply to: triada95@ocsystems.com
> From: J.M.KAMRAD.II@cdev.com (j.m.kamrad.ii)
> To: Multiple recipients of list
> Subject: Re: FREE: AdaSAGE CD-ROM Nov. 95 Edition
> >If your at TriAda 95 just stop by the INEL AdaSAGE booth for a free
> >copy of the Nov. 95 AdaSAGE CD-ROM. Contents include:
> >
> > - AdaSAGE 5.0 with WIN32 support.
> > - Source distributions of AdaSAGE 4.2 for Linux/GNAT
> > and Windows NT/GNAT.
> > - Hypertext, WordPerfect and ASCII documentation versions.
> >
> >In the booth you can see live demos and find out what AdaSAGE is and
> >how it can help you deliver Ada applictions for DOS, Windows 3.1,
> >Windows 95, Windows NT and UNIX systems.
> >
> >Come on by and say HI to Howard, Jon and Dave. Tell them Paul sent
> >you.
>
> Paul,
>
> This sounds great but I won't be attending Tri-Ada. How might I get a copy?
>
> Mike
>
> ------------------------------------
> Mike Kamrad
> Computing Devices International kamrad@cdev.com
> M/S BLC W2J 1.612.921.6908
> 8800 Queen Avenue South FAX: 1.612.921.6552
> Bloomington MN 55431
>
>
>
Well Mike I just sent another posting to the list about that. E-Mail
me if you have any other questions.
From: gse (Scott Evans)
Subject: ADMIN: replies go to the list
Date: Tue, 7 Nov 1995 13:28:02 -0500 (EST)
A quick note for everyone's information: replies to messages on triad95 are
directed to the entire list, to facilitate discussion. If you wish to send
non-public mail in response to something posted here, you'll have to edit
your "To:" line before sending the message.
-----------------------------------------------------------------------------
Scott Evans gse@ocsystems.com
Software Engineer (703) 359-8167
OC Systems, Inc. http://www.ocsystems.com/
-----------------------------------------------------------------------------
From: bowers@krems.kmr.ll.mit.edu (John L Bowers)
Subject: Re: ARA Poster
Date: Wed, 8 Nov 95 07:34:17+120
>The Ada Resource Association is developing a new poster, Ferruccio's "The
Vision
>of Ada." In a few weeks, the image will be available for viewing on the various
>Ada Web pages. At TRI-Ada'95, attendees could request a copy of the poster by
>filling out a card. Readers of these notes can request one by sending e-mail
>with the subject "ARA Poster" to 73313.2671@compuserve.com (that's Bob Mathis
>the Executive Director of the ARA). Please send the usual mailing information
>plus phone, fax, and other relevant contact information.
>
[snip]
>As a separate offer, not available at the Conference, the ARA also has some
>remaining copies of last year's poster showing a surfer. You can also request
>one of these via e-mail. There were some requests last year that were never
>filled by a company no longer working for the ARA. We have recovered their
stock
>and want the opportunity to try again.
>
This sounds great. I would like to request both, since it is not ever
likely I will be sent back to the "world" for Ada anything. I rely on this
and the electronic Ada Digest to stay up to date with what's happening.
Thanks,
John L Bowers
P.O. Box 1158
APO, AP 96555 (That's U.S. Mail only, no special postage required)
------------->
John L Bowers
------------->
Raytheon
Range Systems Engineering (RSE) | Kwajalein Atoll @ E167 43'Long, N08 43'Lat
Radar Dept. / Software Engineer | Republic of the Marshall Islands
By U.S. phone -+- By e-mail
805-238-7994 Ext's 1537/1980/3692 | Internet: BOWERS@KREMS.KMR.LL.MIT.EDU
805-461-7422 Ext's 1537/1980/3692 | Alternate: BOWERS@RAIN.ORG
From: "Paul Whittington"
Subject: FREE AdaSAGE CD
Organization: INEL Special Applications Unit
Date: Tue, 7 Nov 1995 14:24:01 MDT
I've gotten approval to mail CDs to all who E-Mail me with a request
to do so. Please include your complete land address.
I will put you on a list in as recieved order. We will mail to list
members in order until we run out of CDs so get your order in ASAP.
From: "FERGUSON.DOUG-"
Subject: Re: FREE AdaSAGE CD
Date: Tue, 07 Nov 95 17:51:42 EST
I would like a free AdaSAGE CD!
doug
ferguson@sqq89e.bwi.wec.com
______________________________ Reply Separator _________________________________
Subject: FREE AdaSAGE CD
Author: triada95@ocsystems.com at BALT.SMTP
Date: 11/7/95 4:00 PM
I've gotten approval to mail CDs to all who E-Mail me with a request
to do so. Please include your complete land address.
I will put you on a list in as recieved order. We will mail to list
members in order until we run out of CDs so get your order in ASAP.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Ada Policy
Date: 07 Nov 95 22:06:26 EST
Chuck Engle was passing out some small cards describing Ada Policy. I've copied
the contents here because it's a short summary of what's important. The AdaIC
(800-232-4211 or 703-681-2466) has copies if you want one.
Why Ada?
Why is Ada appropriate?
Support for large, complex systems
Interoperability and maintainability
Software Engineering
Modifiable, reliable, portable, easily integrated, etc.
Economics
DoD core competency, lower lifecycle costs
International standard (ANSI, ISO, FIPS)
Only internationally-standardized object-oriented language
Only language with required validation
Promotes reuse, portability
Not locked into proprietary vendor
Most companies settle on a standard, why not DoD
Metrics
60-80% of software costs are in maintenance
Ada best in FAA and SEI scores (capability, cost, risk, etc.)
Ada leads in MITRE reliability and maintainability comparisons
Ada Policies
DODD 3405.1 Ada is the preferred common HOL.
Based on lifecycle cost, prefer use of : (1) COTS and advanced software
technology, when no government modification or maintenance during lifecycle; (2)
Ada; (3) DoD-approved standard HOL, if waiver granted.
Use Ada for all major upgrades (1/3 or more of lines total).
Army extensions: HQDA ltr 25-92-1, 25-95-1
Ada for all modifications of 1/3 or more of functional component.
SQL is approved for DBMSs.
4GLs permitted for prototypes, short-term, ad-hoc systems, non-Ada prototype
cannot be fielded.
Navy extensions: NAVINST5234.2A
Ada for modifications of 1/3 or more of computer software configuration item or
sub-system specification, within 5 years.
Waivers granted only on substantiation of economic analysis.
Air Force extensions: SAF/AQK Action memo
Distinguishes exceptions/waivers, gives details on each.
Exempts individual-use, unique, in-house applications.
SAF/AQK Info Memo Interprets term "cost effective" in Congressional Ada mandate
All three Services permit baselined ("project-validated") compilers - projects
can keep using same compiler throughout lifecycle (after validation certificate
expires).
Ada Information Clearinghouse
800/232-4211 or 703/681-2466
adainfo@sw-eng.falls-church.va.us
URL http://sw-eng.falls-church.va.us
Defense Information Systems Agency, Center for Software
Ada, The Language For a Complex World
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Conference Opening
Date: 07 Nov 95 22:06:30 EST
Hal Hart ran an early session explaining SIGAda to new attendees. I assume most
readers of this list are SIGAda members, if not contact Hal
(halhart@sw-eng.falls-church.va.us) for membership information.
At the main opening session, Hal presented SIGAda awards to Chuck Engle, Mark
Gerhart, Rick Conn, Robert Dewar, Jean Ichbiah, and Tucker Taft.
Silicon Graphics provided some advance displays for the opening plenary session.
It gave the session a big conference, high-tech feel. John Mashey (Silicon
Graphics) gave the opening keynote stressing the trends in technology to larger
and faster computing and how this would change programming and human
interaction. It was a very nice keynote presentation looking toward the future
and opening minds to new approaches. Software innovation will be essential for
bridging the gap between more capabile hardware and pretty much the same wetware
(human beings). E-mail is a very low-bandwidth communication medium and really
inadequate to convey the details, or the content, or the feel of this
presentation. Sorry, you had to be there.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: R&R Compiler
Date: 07 Nov 95 22:06:33 EST
R&R Software was giving away some free copies of their Ada95 compiler for
Windows NT or Windows 95. They were also offering "rain checks" so that people
could get a copy by sending them $20 for postage and handling. The offer expires
December 15. The address for R&R Software is P.O.Box 1512, Madison, WI 53701.
Their phone and FAX are 608-251-3133 and 608-251-3340, respectively, but send
them a check, don't call.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Job Openings
Date: 07 Nov 95 22:06:38 EST
Patrick McDermott (a technical recruiter from Phoenix, AZ) (qrp@aol.com) said he
had over 450 job listings for Ada programmers. These are sales openings,
premanent progamming openings, consulting, and so forth. The word needs to get
out that there are Ada jobs out there.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Validations
Date: 07 Nov 95 22:06:36 EST
Intermetrics announced the first Ada95 validation two weeks ago.
This week ACT and SGI are validating on the exhibit floor. They are passing all
of the core tests, all of the annex tests, and almost all of the optional tests.
They are validating on three different platforms - Indy, Indigo-2, and Onyx.
Results of the validation should be available and announced later in the week.
OC Systems is 98 and 44/100s complete. That is the exact percentage as of last
week. They would probably have finished except for taking a break to attend this
conference.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Searching for Ironman
Date: 07 Nov 95 22:06:41 EST
I had the opportunity to talk with Bill Whitaker at TRI-Ada'95 about the
following message that originally appeared on the Team-Ada list about a week
ago.
Summarizing Bill's comments: C was not formally evaluated against Tinman in the
late 70's. DARPA had a special working relationship with AT&T Bell Labs at the
time because of UNIX. Bells Labs offered to sell UNIX to DARPA for $15,000 (late
70's dollars, that was for everything). They tried to get the "fathers" of all
the existing languages to participate in the proposals and evaluations. Bell
Labs responded that C did not meet the requirements (in particular the
requirements for readability and security) and they didn't want to submit it for
evaluation. In the early 80's, people at AT&T were working on an Ada compiler
and those people were located physically near the C++ team. According to Bill,
AT&T lost a major (multiple 100 millions of dollars) because of their inability
to develop a validated Ada compiler in the proper time frame. C++ was clearly
influenced by the design and implementation of Ada.
One of the advantages of attending a conference like TRI-Ada'95 is to meet and
talk with some of the central people in the Ada program.
-- Bob Mathis
- - - - - - referenced message - - - - - - - - - - - - - -
- -
I am working on a research paper for a course in "Programming Paradigms". The
topic is the "Ada Mandate, is it still valid?" I intend to bounce the draft of
the C++ standard against the original requirements for Ada, the
Ironman document. I have searched for it but have been unable to locate an
elecronic copy, if it exists. Does anyone out there have a copy or a pointer to
one? I prefer not to chase it through DTIC :-)
__________________________________________________
Daniel McDonough mcdan@rt66.com
Team Ada
Team OS/2
__________________________________________________
From: nicolas pesenti
Subject: Re: FREE AdaSAGE CD
Date: Wed, 8 Nov 95 7:55:36 MET
Yes, I would appreciate receiving the AdaSAGE CD.
Here is my postal address:
Nicolas PESENTI
EINEV
1, rte de Cheseaux
CH-1400 Yverdon
Switzerland
Thanks in advance,
Nicolas
>
> I've gotten approval to mail CDs to all who E-Mail me with a request
> to do so. Please include your complete land address.
>
> I will put you on a list in as recieved order. We will mail to list
> members in order until we run out of CDs so get your order in ASAP.
>
>
From: Peter.Hermann@csv.ICA.Uni-Stuttgart.DE (Peter Hermann)
Subject: Re: R&R Compiler
Date: Wed, 8 Nov 1995 11:14:12 +0100 (MET)
>
> R&R Software was giving away some free copies of their Ada95 compiler for
> Windows NT or Windows 95. They were also offering "rain checks" so that people
> could get a copy by sending them $20 for postage and handling. The offer expires
> December 15. The address for R&R Software is P.O.Box 1512, Madison, WI 53701.
> Their phone and FAX are 608-251-3133 and 608-251-3340, respectively, but send
> them a check, don't call.
do they have e-mail? (this may be a question of general interest)
--
Peter Hermann Tel:+49-711-685-3611 Fax:3758 ph@csv.ica.uni-stuttgart.de
Pfaffenwaldring 27, 70569 Stuttgart Uni Computeranwendungen
Team Ada: "C'mon people let the world begin" (Paul McCartney)
Subject: Re: Validations
From: Cheryl Marquis
Date: Wed, 08 Nov 1995 07:34:30 -0500
> Intermetrics announced the first Ada95 validation two weeks ago.
>
> This week ACT and SGI are validating on the exhibit floor. They are passing all
> of the core tests, all of the annex tests, and almost all of the optional tests.
> They are validating on three different platforms - Indy, Indigo-2, and Onyx.
> Results of the validation should be available and announced later in the week.
>
> OC Systems is 98 and 44/100s complete. That is the exact percentage as of last
> week. They would probably have finished except for taking a break to attend this
> conference.
>
Good info coming off this mailing list!!! Wish I could be there, but thanks
for keeping us posted! I've already passed this one around to everyone I work
with.
Cheryl
From: "Karen J. Christ 516-851-6103"
Subject: Re: FREE AdaSAGE CD
Date: Wed, 08 Nov 1995 09:10:41 -0500 (EST)
> I've gotten approval to mail CDs to all who E-Mail me with a request
> to do so. Please include your complete land address.
>
> I will put you on a list in as recieved order. We will mail to list
> members in order until we run out of CDs so get your order in ASAP.
Please add my name to the list. Thanks in advance.
Karen J. Christ
Reuters Information Technology
88 Parkway Drive South
Hauppauge, NY 11788
E-mail: karen.christ@reuters.com
From: "Coniam, Todd"
Subject: RE: FREE AdaSAGE CD
Date: Wed, 08 Nov 95 08:12:00 CST
Like many in this world, my e-mail system strips off most header info.
In the future everyone, please remember to use a signiture line so
that we can know who to send private e-mail to without having to
clutter the mailing list.
Please send a copy of the AdaSAGE CD to me at my address below.
Todd Coniam, MSgt, USAF |
Chief, Database Software Development | Hq SSG OL-B/SDCB
WWOLS Replacement Project | 3580 D Avenue
EMail: coniam@comswsys.tinkernet.af.mil | Tinker AFB, OK 73145-9155
Phone: 1-405-734-3283 ext. 23 |
DSN: 884-3283 ext. 23 | FAX: 1-405-734-7302/4373
----------
From: triada95
Subject: FREE AdaSAGE CD
Date: Tuesday, 07 November, 1995 12:27
I've gotten approval to mail CDs to all who E-Mail me with a request
to do so. Please include your complete land address.
I will put you on a list in as recieved order. We will mail to list
members in order until we run out of CDs so get your order in ASAP.
From: Sundog Software
Subject: Re: e-mail for R&R Compiler
Date: Wed, 8 Nov 1995 08:19:44 -0600 (CST)
Respond to rBrukardt@Bix.Com. I also think they [will] have a web page
on ASEET where you can request info or place orders.
Sundog.Software@Msn.FullFeed.Com
From: gse (Scott Evans)
Subject: ADMIN: AdaSage offer
Date: Wed, 8 Nov 1995 10:34:50 -0500 (EST)
Please, folks, if you want to request an AdaSage CD, send mail directly
to Paul Whittington at phw@inel.gov, rather than to the whole list.
Thanks!
-----------------------------------------------------------------------------
Scott Evans gse@ocsystems.com
Software Engineer (703) 359-8167
OC Systems, Inc. http://www.ocsystems.com/
-----------------------------------------------------------------------------
From: theeke@reuse.asset.com (Patrick A. Theeke)
Subject: Re: e-mail for R&R Compiler
Date: Wed, 8 Nov 1995 12:22:32 -0500 (EST)
>Respond to rBrukardt@Bix.Com. I also think they [will] have a web page
>on ASEET where you can request info or place orders.
There is indeed a page for RR Software at ASSET. See
http://source.asset.com/rrsoft.html
for their WWW pages and links to the product description and ordering
pages.
--
| Patrick A. Theeke
_/|__ _ Deputy for Software Engineering/Senior Software Engineer
_/ * |/_\ Asset Source for Software Engineering Technology (ASSET)/SAIC
/ _/ 1350 Earl L. Core Road P.O. Box 3305 VOICE: (304) 284-9000
( / Morgantown, WV 26505 FAX: (304) 284-9001
\____/ theeke@source.ASSET.com http://source.asset.com/
From: Sundog Software
Subject: FREE Adasage CD-Rom
Date: Wed, 8 Nov 1995 11:24:04 -0600 (CST)
From: "Paul Whittington"
>how it can help you deliver Ada applictions for DOS, Windows 3.1,
>Windows 95, Windows NT and UNIX systems.
One assumes you have given up on the RR version (no one's heard from you
since last TriAda if I remember correctly) ?
> - Source distributions of AdaSAGE 4.2 for Linux/GNAT
> and Windows NT/GNAT.
Is this sufficent source that I could attempt the port to Janus/Ada for
you ?
If not a problem could you drop a CD off at RRS booth for me or you could
mail me a copy :
Isaac Pentinmaki 3446 Prairie Dr. Deerfield WI 53531
From: "Paul Whittington"
Subject: Re: FREE Adasage CD-Rom
Organization: INEL Special Applications Unit
Date: Wed, 8 Nov 1995 11:55:50 MDT
> Date sent: Wed, 8 Nov 1995 10:19:24 -0800
> Send reply to: triada95@ocsystems.com
> From: Sundog Software
> To: Multiple recipients of list
> Subject: FREE Adasage CD-Rom
> From: "Paul Whittington"
>
>
> >how it can help you deliver Ada applictions for DOS, Windows 3.1,
> >Windows 95, Windows NT and UNIX systems.
>
> One assumes you have given up on the RR version (no one's heard from you
> since last TriAda if I remember correctly) ?
>
> > - Source distributions of AdaSAGE 4.2 for Linux/GNAT
> > and Windows NT/GNAT.
>
> Is this sufficent source that I could attempt the port to Janus/Ada for
> you ?
>
> If not a problem could you drop a CD off at RRS booth for me or you could
> mail me a copy :
>
> Isaac Pentinmaki 3446 Prairie Dr. Deerfield WI 53531
>
>
Hello Isaac,
Here at the U.S. Dept. of Energy's Idaho National Engineering
Laboratory (INEL) we work on a cost recovery basis. The bottom line
is that we can only port to and maintain ports for which we can
obtain a sponsor (read financial source). To date no sponsor has
stepped forward and offered to fund a port to the RR Ada 95 compiler.
Frankly we would love to do the port and maintain it. Please let us
know ASAP if you are aware of such a sponsor.
It is true that we started to port AdaSAGE to RR's Ada 95 compiler
last year. This was an AJPO sponsered effort to get some experience
with Ada 95, and find out how difficult it would be to port to Ada 95.
At the time RR's compiler was in early beta and had some limitations
that did not allow us to finish the port.
The source code provided on the CD is sufficent to port AdaSAGE
version 4.2 and all of its associated tools to the JANUS compiler,
and if you would like to do it GREAT! Let us know what we can do to
help. We would also be happy to make the port available on our FTP
site and on our next CD. For some time now we have been trying to
figure out how to work with Ada compiler vendors through the DoE to
get the vendors access to the source to AdaSAGE, under NDA, so that
they might do ports and maintenance. I suggest that you call Bobbi
Smith, here at the INEL, at (208) 526-0763 if you are interested.
Call next week as she is at TriAda this week.
Hey Isaac, if you are at TriAda just walk accross the hall from your
booth to the INEL booth and pick up a CD, otherwise make sure one of
your people picks one up. If you don't get one at the conference let
me know and we'll get one to you.
Keep up the good work, we need Ada 95 based software development
products that can compete with Delphi, VC++, VB 4.0, BC++ etc.
TTFN Paul
From: Chris.Morgan@baesema.co.uk
Subject: help!
Date: Wed, 8 Nov 1995 17:05 GMT
Dear all,
Due to local problems I missed all messages to the triada95 distribution list
until lunchtime GMT today, Wednesday (my subscription didn't get out of our
network). What goodies did I miss? I'm interested in GNAT and other Ada95
compilers in particular. I'm subscribing on behalf of approx 20 others besides
myself so I'd really appreciate any forwarded previous postings.
This list is a great idea by the way.
Regards,
Chris
-------------------------------------------------------------------------------
-- Chris Morgan. BAeSEMA Ltd, Scientific House,
-- 40-44 Coombe Rd, New Malden, Surrey, UK.
-- Phone (UK) 0181-942-9661, Fax 0181-949-8067
--
-- E-mail : chris.morgan@baesema.co.uk
--
-- Currently grappling with Ada95 via GNAT 2.07 on Sun-Sparc-Solaris2.4
--
-------------------------------------------------------------------------------
-- Team Ada
-------------------------------------------------------------------------------
From: "Guion, Martin"
Subject: Re: FREE Adasage CD-Rom
Date: Wed, 08 Nov 95 15:45:00 CST
I would like the free CD-Rom.
My address is:
Martin Guion
3281 Ridgecrest Ct. Apt. 1011
Norman, OK 73072
Thanks
guion
----------
From: triada95
Subject: Re: FREE Adasage CD-Rom
Date: Wednesday, November 08, 1995 10:52AM
> Date sent: Wed, 8 Nov 1995 10:19:24 -0800
> Send reply to: triada95@ocsystems.com
> From: Sundog Software
> To: Multiple recipients of list
> Subject: FREE Adasage CD-Rom
> From: "Paul Whittington"
>
>
> >how it can help you deliver Ada applictions for DOS, Windows 3.1,
> >Windows 95, Windows NT and UNIX systems.
>
> One assumes you have given up on the RR version (no one's heard from you
> since last TriAda if I remember correctly) ?
>
> > - Source distributions of AdaSAGE 4.2 for Linux/GNAT
> > and Windows NT/GNAT.
>
> Is this sufficent source that I could attempt the port to Janus/Ada for
> you ?
>
> If not a problem could you drop a CD off at RRS booth for me or you could
> mail me a copy :
>
> Isaac Pentinmaki 3446 Prairie Dr. Deerfield WI 53531
>
>
Hello Isaac,
Here at the U.S. Dept. of Energy's Idaho National Engineering
Laboratory (INEL) we work on a cost recovery basis. The bottom line
is that we can only port to and maintain ports for which we can
obtain a sponsor (read financial source). To date no sponsor has
stepped forward and offered to fund a port to the RR Ada 95 compiler.
Frankly we would love to do the port and maintain it. Please let us
know ASAP if you are aware of such a sponsor.
It is true that we started to port AdaSAGE to RR's Ada 95 compiler
last year. This was an AJPO sponsered effort to get some experience
with Ada 95, and find out how difficult it would be to port to Ada 95.
At the time RR's compiler was in early beta and had some limitations
that did not allow us to finish the port.
The source code provided on the CD is sufficent to port AdaSAGE
version 4.2 and all of its associated tools to the JANUS compiler,
and if you would like to do it GREAT! Let us know what we can do to
help. We would also be happy to make the port available on our FTP
site and on our next CD. For some time now we have been trying to
figure out how to work with Ada compiler vendors through the DoE to
get the vendors access to the source to AdaSAGE, under NDA, so that
they might do ports and maintenance. I suggest that you call Bobbi
Smith, here at the INEL, at (208) 526-0763 if you are interested.
Call next week as she is at TriAda this week.
Hey Isaac, if you are at TriAda just walk accross the hall from your
booth to the INEL booth and pick up a CD, otherwise make sure one of
your people picks one up. If you don't get one at the conference let
me know and we'll get one to you.
Keep up the good work, we need Ada 95 based software development
products that can compete with Delphi, VC++, VB 4.0, BC++ etc.
TTFN Paul
From: FRAVEL@aaicorp.com
Subject: Ada Policy -Reply
Date: Wed, 08 Nov 1995 17:19:55 -0500
I have a couple of policy questions that should be answerable from the
multitude of experts at TriAda.
1. Where is the official policy on Ada'95? Everything I have or can pull
from Ada-IC still references only Ada'83 (i.e., 1815).
2. I f I have a microprocessor/microcontroller that is the best engineering
choice (in an embedded MCCR application) but for which there is no Ada
compiler, do I need a waiver? This is for a joint AF/Navy program.
Thanks - Bill
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Lloyd Mosemann
Date: 08 Nov 95 21:47:52 EST
At the government track, Lloyd Mosemann gave a brief introduction. Mr. Mosemann
(UnderSecretray of the Air Force) has been a very strong supporter of Ada. He
has announced his retirement in early January, 1996. He expressed surprise that
he didn't sense more enthusiasm at the conference, because he has been seeing
very good results with using Ada and modern software engineering techniques
(including particularly reuse and product line engineering).
He thinks that "best commercial practices" involves a move away from individual
programmer craftmanship ("hacking") thoward product lines and management vision.
Learning Ada and software engineering occurs at many levels. Ada should be the
natural language of choice. He thinks the "mandate" is on the way out, but the
use of Ada should increase. "Ada is the answer for this nation in this era of
competitiveness, but the Ada industry has to help make it happen."
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Vendor Announcements
Date: 08 Nov 95 21:47:56 EST
TLD plans to provide products based on GNAT and GNU technology. They have a
strong record of customer support in cross embedded environments and think this
will provide a cost effective way to help their customers migrate to Ada95. TLD
believes this will enable them leverage their expertise and software into an
increased marketplace without the expense of original development. They also
intend to continue to support and enhance their existing Ada83 products. For
more inromation contact them at tld@cerf.net.
Irvine Compiler is migrating their existing compiler to Ada95. They have
released the i960MX compiler with 33 bit secure systems features. Joe Kohli has
been traveling the world promoting this compiler. Security features are enforced
by hardware. Contact them at JKohli@Irvine.com.
Kathleen Gilroy, Software Compositions, was there promoting their work in
quality assessment and improvement, reengineering for reuse, and Ada83 to Ada95
transition. Contact her at gilroyk@source.asset.com.
DDC-I has been having a good time at the conference. Their booth is next to
Rational's at the entrance. They told me it was their overflow that was keeping
Rational so busy. (Rational, of course, told me it was the other way around.)
DDC-I anounced Ada compilers for Windows NT. During 1996 they will be migrating
these to Ada95. DDC-I also announced their Sun SPARC Ada95 native compiler,
which they expect to validate in December. Contact Jennifer Sanchez or visit
their Web page at http://www.ddci.dk/ddci/.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Environment Initiatives
Date: 08 Nov 95 21:48:59 EST
Chuck Engle introduced a panel discussing how to avoid the mistakes of
introducing Ada83.
The validation procedures have been evolved to focus somewhat more of meeting
user needs. The test suite has been improved to replace incorrect tests more
quickly. All the Ada vendors are well aware of these changes. The general hope
is this will make Ada95 validations more relevant to using the language.
Ada95 compilers and programmers are evolving more quickly since they can build
on existing Ada83 knowledge.
Ada is not a silver bullet. Good programmers with knowledge of the application
domain are necessary. Existing Ada83 compilers are usually old technology, the
important thing is to look at the new Ada environments which are competitive
with other programming environments.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Stallman's Wednesday Plenary
Date: 08 Nov 95 22:07:28 EST
Richard Stallman talked about the historical background and philosophy of the
Free Software Foundation. Free means freely usable and redistributable under the
"copy-left" arrangement, not necessarily free of cost. GNU and GCC served as the
underlying framework for GNAT.
It was an interesting, but rambling talk. GNAT has changed the way people think
about the availability of Ada compilers. It was very useful to have Stallman
himself describe his philosophy, which he thinks of as a moral position. Robert
Dewar, head of the GNAT project, has restated and explained this philosophy
frequently in various Ada forums including a local LA SIGAda talk the night
before.
Stallman expressed a general approval for the changes made in Ada95. He would
have made overload resolution less dependent on the context and redefinition of
functions on tagged types require a more explicit declaration of intent by the
programmer, but he didn't consider these major.
At the end of the talk, there was a discussion of potential changes in copyright
law which would have a negative effect on free software and other intellectual
property. He pointed to an upcoming article in the January issue of Wired.
The funny thing that occured was the difficulty the hotel had in supplying him
with a cup of hot tea.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Steve Zeigler's study of Ada and C costs
Date: 08 Nov 95 22:07:22 EST
I wasn't able to go to all the sessions. This is one in particular I'm sorry I
missed. Norm Cohen gave me the following report. I hope Steve and Rational make
more of this information available.
For years, Ada advocates have preached the benefits of Ada for increasing
program reliability and reducing life-cycle costs. For anyone who has
programmed in both Ada and C, the advantages of Ada are intuitively obvious, but
when challenged to provide hard empirical evidence of these advantages, we have
been tongue-tied.
Steve Zeigler's presentation, describing the experience of Verdix (now part of
Rational) in developing and maintaining its code base--which is about equally
divided between Ada and C code--provides compelling evidence that belongs in the
arsenal of every Ada advocate.
Steve has been collecting data for a number of years and has anaylzed it
carefully to discover patterns. Here are some of the most striking results he
presented:
- Ada code required 4.59 fixes per thousand lines of code, compared with 9.21
for C.
- Ada code contained .096 customer-reported defects per thousand lines of code,
compared with .676 for C.
- Development costs for Ada code were $6.62 per noncomment/nonblank line,
compared with $10.52 for C.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: SIGAda Extended Executive Committee
Date: 08 Nov 95 22:06:58 EST
On Tuesday night, I attended the SIGAda Extended Executive Committee meeting.
There were a couple of nice hospitality suites that I didn't go to. Readers may
have noticed that I haven't been able to attend all the sessions and that my
reports so far have not covered all aspects of the conference.
The SIGAda budget has been adjusted to reflect changes in membership levels and
TRI-Ada income. SIGAda membership has not declined as much as for SIGs in
general. Attendance at this TRI-Ada is very close to projections and expenses
have been controlled very well, so the expected surplus from the conference
should be realized. (This is subject to the usual accounting disclaimers.)
There was a review of SIGAda working group activities. Some of these groups had
meetings at other times during the conference and reports have been (are being)
posted. Some of the working groups are planning activities in conjunction with
the AdaEurope meeting, June 10-14, 1996, in Montreux, Switzerland. There may
also be some meetings in conjunction with STC, Salt Lake City, April 22-26,1996.
After three hours of other discussion (at 2:30am EST), we finally came to the
topic of next year's meeting. It is unusual for the next year's TRI-Ada to not
be decided before the current year TRI-Ada starts. A proposal was made to hold
next year's conference in Montreal, Canada. This was eventually voted down (at
about 4:00am EST), but that left SIGAda without a plan for next year. (I've been
involved with SIGAda and its predecessor organizations since the time when
meetings might include us all going to the same restaurant for lunch at the same
time. The current model of a single big conference and tradeshow is broken. The
challenge for SIGAda and the Ada community is to come up with a new vision for
meetings and communication.)
Mime-Version: 1.0
From: mcdan@rt66.com
Date: Wed, 08 Nov 95 21:14:22 +0000
Thanks to all who responded to my request. I found a copy of
Steelman locally.
There was some confusion about what I am doing and I
would like to see some discussion about it.
I have been in a running battle for a couple of years with
another organization about the use of Ada vs C++. This
other organization has spent $$$$ recreating software in C++
that exists and is running in Ada. My intent with this report
I am writing for a CS class is to:
1. See if the _Current_ Draft of _C++_ meets the original
requirements that led to the design of Ada. (If not, then
how do you justify using C++ instead of Ada)
2. Discuss the conditions of the "Software Crisis" that led
the DoD to head to a Common Programming Language. (Will use
of C++ open the door to language proliferation?)
3. Discuss the current views of the Software Engineering
community. (C++ and programming in the large, does it
work?)
While this other organization recently became "a team player",
I expect this battle to be rejoined shortly. I can not fight
this battle against managers that have religous
zealots advising them unless I have a solid grasp of the facts and
can prove that I know what I am talking about.
Irrespective of the merits of the Ada Mandate, It is the LAW, the DoD
has policy implementing it, with proceedures to get a waiver. It
burns my wallet as a taxpayer to see an organizations "SPOOKIER THAN
THOU" attitude allow them to waste $$$$ duplicating another projects
efforts, ignore higher headquarters policy, violate the law, and get
away with it.
COMMENT?
PS Wish I was at TRI-Ada
PPS If this is a duplicate post - Sorry, New mailer
----------------------------------------------------------------------
Daniel McDonough mcdan@rt66.com
Team Ada
Team OS/2
From: Chris.Morgan@baesema.co.uk
Subject: RE: Stallman's Wednesday Plenary
Date: Thu, 9 Nov 1995 12:08 GMT
Bob Mathis wrote (with deletions) :
>It was an interesting, but rambling talk. GNAT has changed the way people think
>about the availability of Ada compilers. It was very useful to have Stallman
>himself describe his philosophy, which he thinks of as a moral position. Robert
>Dewar, head of the GNAT project, has restated and explained this philosophy
>frequently in various Ada forums including a local LA SIGAda talk the night
>before.
>
The GNAT library model, a Stallman insight, has also changed the way people
*implement* Ada compilers. I believe AdaMagic is source-based as well as GNAT.
Comments?
Chris
-------------------------------------------------------------------------------
-- Chris Morgan. BAeSEMA Ltd, Scientific House,
-- 40-44 Coombe Rd, New Malden, Surrey, UK.
-- Phone (UK) 0181-942-9661, Fax 0181-949-8067
--
-- E-mail : chris.morgan@baesema.co.uk
--
-- Currently grappling with Ada95 via GNAT 2.07 on Sun-Sparc-Solaris2.4
--
-------------------------------------------------------------------------------
-- Team Ada
-------------------------------------------------------------------------------
From: brashepw@ss2.sews.wpafb.af.mil (PHILIP W. BRASHEAR)
Subject: Re: Ada Policy -Reply
Date: Thu, 9 Nov 95 07:55:53 EST
> I f I have a microprocessor/microcontroller that is the best engineering
> choice (in an embedded MCCR application) but for which there is no Ada
> compiler, do I need a waiver?
Personally, speaking for no one else, I would say "Yes".
On the other hand, if there is no Ada compiler for the processor, I would ask
whether the existence of the best system software (i.e., an Ada compiler)
was considered as part of the "best engineering choice". Software first? (:-)
Phil Brashear
From: westley@hercules.calspan.com (Terry J. Westley)
Subject: Re: Lloyd Mosemann
Date: Thu, 9 Nov 1995 08:47:44 -0500 (EST)
Robert Mathis reports that Lloyd Moseman "expressed surprise that
he didn't sense more enthusiasm at the conference."
By "enthusiasm," do you think he meant attendance or attitude, or
perhaps both?
--
Terry "if I were there, I'd be enthusiastic!!!" Westley
Principal Computer Scientist
Calspan SRL, P.O. Box 400, Buffalo, NY 14225
westley@calspan.com http://worf-gw.calspan.com/~westley/
From: westley@hercules.calspan.com (Terry J. Westley)
Subject: Re: SIGAda Extended Executive Committee
Date: Thu, 9 Nov 1995 09:00:00 -0500 (EST)
Robert Mathis says that we may have noticed that he hasn't attended
all the sessions and so his reports don't cover the whole conference.
I thank and applaud him for his time and effort in making his thoughts
and impressions available to us through this mailing list. Partial
coverage of the conference in this way, for those of us who could not
attend, is greatly appreciated.
Thanks also to OC Systems for sponsoring the mailing list.
--
Terry J. Westley, Principal Computer Scientist
Calspan SRL, P.O. Box 400, Buffalo, NY 14225
westley@calspan.com http://worf-gw.calspan.com/~westley/
From: John Howard
Subject: RE: Stallman's Wednesday Plenary
Date: Thu, 9 Nov 1995 08:20:05 -0600 (CST)
On Thu, 9 Nov 1995 Chris.Morgan@baesema.co.uk wrote:
> Bob Mathis wrote (with deletions) :
>
>>It was an interesting, but rambling talk. GNAT has changed the way
>>people think about the availability of Ada compilers. It was very
>>useful to have Stallman himself describe his philosophy, which he
>>thinks of as a moral position. Robert Dewar, head of the GNAT project,
>>has restated and explained this philosophy frequently in various Ada
>>forums including a local LA SIGAda talk the night before.
>
> The GNAT library model, a Stallman insight, has also changed the way
> people *implement* Ada compilers. I believe AdaMagic is source-based as
> well as GNAT. Comments?
>
> Chris
Most Forth systems have been similarly source-based since 1980. Forth
has a Dictionary structure consisting of child Vocabularies. It is
attributable to the inventor of Forth, Charles H. Moore (founder of FORTH
Inc.) now with Computer Cowboys Inc. He also created the concept of a
defining word that defines other words (i.e. generic template). And he
created Forth having the distinction between compile-time action and
execute-time action. That concept is similar to semantic tree
elaboration. The Forth inner/outer interpreters are the syntax parser
for the language.
I am still learning about Ada95 but increasingly I appreciate the
parallel concepts seemingly derived from Forth and object-oriented
Pascal. Though I would like to see the core ANS Forth become a universal
language to replace assembler languages. Even via a pragma within
Ada95. Ada95 provides a genuinely standard core which is an improvement
over any one particular language that I have used. Ada95 integrates
several useful features from my past favorite languages.
See BYTE magazine (AUGUST 1980 Vol.5;No.8 published by McGraw-Hill) for
"The Evolution of Forth, An Unusual Language" by Charles H. Moore.
-- John Howard -- Team Ada -- Team OS/2
-- jhoward@sky.net
-- Free GNU-based Ada 95 compilers are at ftp://cs.nyu.edu/pub/gnat/
-----------------------------------------------------------------------
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Re: help!
Date: 09 Nov 95 10:13:00 EST
Chris Morgan asked "What goodies did I miss?"
I'm at the conference and keep asking other people that as well.
I hope others will post to this list even after they get home. In
particular, I plan to summarize things on Thursday and Friday and
then post some messages that repeat the important points.
GNAT is validating on some SGI machines at the show. Intermetrics
has already validated a compiler. OC Systems and DDC-I are very close.
Rational is migrating their environment to Ada95. Thomson has teamed
Intermetrics to produce an academic oriented Ada95 compiler. R&R was
distributing sample copies of their Ada95 compiler. Tartan, Digital,
TLD, Irvine, and others also had announcements about exciting products
and migration to Ada95.
There is a lot going on in the Ada industry and at this conference.
The above paragraph is uneven in how different companies are described
and there are other companies I haven't had a chance to talk to yet.
I really want others to post to this list to describe what they're
doing.
"This list is a great idea by the way." Thanks
-- Bob Mathis
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: GNAT validation
Date: 09 Nov 95 10:13:08 EST
Robert Dewar posted the following message to
comp.lang.ada. Since it relates to this conference, he should have also
posted on this list. -- Bob Mathis
'How's that GNAT validation coming along anyway"
we are currently running the vaidation here on the floor at Tri-Ada. If all
goes as planned, we should be validated by Thursday on SGI Onyx, SGI Indy
and SGI Indigo.
The validation profile is that (of course) we pass 100% of the required B
and C tests. In addition, we are passing all the executable tests in all
the optional annex tests, and about 95% of the optional executable tests in
the core ( we pass over 90% of the optional B tests, I am not quite sure
what the exact figure is there). THe remaining core C tests (13 of them
at the current count, but going down all the time, will all be passing
correctly in the product version of the validated compiler.
From: FRAVEL@aaicorp.com
Subject: Re: Ada Policy -Reply -Reply
Date: Thu, 09 Nov 1995 16:01:03 -0500
>>>>>>>>>>>>>>>
> I f I have a microprocessor/microcontroller that is the best engineering
> choice (in an embedded MCCR application) but for which there is no
Ada
> compiler, do I need a waiver?
Personally, speaking for no one else, I would say "Yes".
On the other hand, if there is no Ada compiler for the processor, I would
ask whether the existence of the best system software (i.e., an Ada
compiler) was considered as part of the "best engineering choice".
Software first? (:-)
Phil Brashear
<<<<<<<<<<<<<<<
Phil - in most circumstances I always tend to think "software first".
However, the reasons for "software first" probably do not apply in this
instance. This microcontroller is a replacement for a hardware
component that monitors line power and provides a shut down signal to
other system if the input is out of a pre-set (by specification) range for
several parameters. There is no need for re-programming, the application
is not in any way complex, it's more like a process control application and
I cannot see any life-cycle for this part.
BTW, someone else referred me to the AF directive that interprets the
Ada "mandate/policy" and there is a specific exception for
microcontrollers and microprocessors in some industrial process control
applications.
A waiver for "life cycle cost" is difficult in this situation, as the
comparison basis is lacking. It seems to me that not providing an
exception mechanism for such applications is another reason why we
have so much bad press about Ada. Examples like this tend to get cited
as an unrealistic approach associated with Ada.
Bill
From: brashepw@ss2.sews.wpafb.af.mil (PHILIP W. BRASHEAR)
Subject: Re: Ada Policy -Reply -Reply
Date: Fri, 10 Nov 95 07:06:20 EST
Bill: "Examples like this tend to get cited as an unrealistic approach
associated with Ada."
MEA CULPA! "Unrealistic approach" could often be accurately applied to
my opinions. I'm an idealist, and more than a bit paranoid. I think
you're right in this case.
(Oops! That means I'm wr___. Well, I don't guess it's the first time.)
Thanks,
Phil
From: Rush Kester
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 10 Nov 1995 13:42:50 -0500 (EST)
On Thu, 9 Nov 1995 FRAVEL@aaicorp.com (with my deletions) wrote:
> > If I have a microprocessor/microcontroller that is the best engineering
> > choice (in an embedded MCCR application) but for which there is no Ada
> > compiler, do I need a waiver?
Bill,
You (as well as others in a similar situation) may want to consider
using GNAT to creat a cross-compiler for the microcontroller you are
using.
I understand, from Robert Dewar's talk (or side discussions) at the most
recent joint meeting of Baltimore & Washington D.C. SIGAda's that a good
bit of the code generation process is table driven and consists of
describing the target machine (# registers, etc.) If you (or your
project/company/customer) prefer a COTS solution, you should contact ACT
for a quote.
Rush Kester
Team-Ada
From: dprice@powergrid.electriciti.com (David A Price)
Subject: Re: Lloyd Mosemann
Date: Fri, 10 Nov 95 18:12 PST
The term "hacking"
originates back when the solution to a model
train track switch was done in the most optimized way,
thus considered a "good hack" on the track system.
To much process and not enough product has been Ada's downfall.
dp
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Bests Awards
Date: 11 Nov 95 00:57:50 EST
There were a number of excellent nominees
Best Paper: "DVM: An Object-Oriented Framework for Building Large Distributed
Ada Systems," by Christopher J. Thompson (Hughes Aircraft) and Vincent Cellier
(Hughes Canada)
Best Presentation: "Ada95 as Implementation Language for Object-Oriented
Designs," by Stephane Barbey (Swiss Federal Institute of Technology)
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Microsoft and Ada
Date: 11 Nov 95 00:58:01 EST
Chuck Engle announced at the closing plenary, that he had received official
notice from Microsoft that they have accepted our Ada95-Win32 bindings and will
be distributing them from their server. It is not a Microsoft product, but it is
the first non-Microsoft language product they have endorsed.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Dave Emery Comments
Date: 11 Nov 95 00:58:18 EST
Dave Emery had the general comment that the papers at the conference were very
good overall (just a little plug for ACM, the conference proceedings are
purchasable separately). GNAT has triggered lots of exciting work. EPFL (Swiss
Fed. Inst. of Tech.) continues to lead in object-oriented programming insights.
More than 50% of "Best Paper" nominees were European.
Arne Carlsson (Saab Ericcson Space) described a satellite system using
RAD-hardened microprocessor with Ada-specific instruction set. He also asks why
aren't there more Ada processors? We couldn't get performance under space, size,
power, RAD, etc. constraints without hardware support for the language (for any
language)
p.s. Emery denies that his finger injury is due to strong typing or punching C
programmers...
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Closing Plenary
Date: 11 Nov 95 00:58:06 EST
Peter Coffee, PCWeek, talked about an Ada program he had done to exercise the
Pentium division bug about a year ago. He had to get it working by his deadline,
which is a real time issue in his context. He also discussed the difficulty
others had in translating his program into C.
In every product development, there is a time when software is on the critical
path. He gave different examples of where being first with a particular software
based product gave companies a step up in their market. Project abandonment is a
cost in software development that is just beginning to be recognized.
He talked about requirements for high reliable microprocessor software used in
non-computer appliances. He said that when his camera wasn't focusing properly,
knowing the software patch was on CompuServe was no help. The fraction of
software that goes into desktop applications is very small.
Peter has very positive feelings about the usefulness of Ada and concerns about
the popularity of C++; but Ada is being mentioned less and less in magazine
articles. He mentioned John McCormick's positive experience using Ada in a
project course. He encouraged Ada people to be more active in publicizing their
successes.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: General Summary
Date: 11 Nov 95 01:02:18 EST
My own personal conclusion at the end of the conference was very positive. This
is much better than the ending feeling of some other recent Ada conferences.
Terry Doran, Jim Moore, and other members of the committee deserve our thanks
and congratulations. Job well done.
On the other hand, I think the current TRI-Ada model needs considerable work.
SIGAda is planning the next big conference for December 1-6, 1996, in
Philadelphia. Comments and suggestions are welcome. This list
(TriAda95@ocsystems.com) is a good place to discuss TRI-Ada'95, reactions to it,
and suggestions for next year. After this list goes away (November 30, 1996),
the ARA list (ara@ocsystems.com) is an appropriate place to discuss Ada
conferences.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: C++, Smalltalk, Ada Panel
Date: 11 Nov 95 01:01:44 EST
The final session was a panel comparing Ada, C++, and Smalltalk. There were
three good proponents of the languages who stressed similar goals - reuse,
reliability, information hiding, flexibility, and so on. Mostly their choices
were based on their own priorities, but they understood how others might have
different choices.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: TRI-Ada'96
Date: 11 Nov 95 01:02:11 EST
TRI-Ada'96 will be held in Philadelphia, PA, December 1-6, 1996. Philadelphia is
emphasizing 1996 as the "Year of the Computer" in honor of the 50th anniversary
of the ENIAC at the University of Pennsylvania.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Why we don't use Ada panel
Date: 11 Nov 95 01:01:52 EST
There were three main speakers in a panel on "Why we don't use Ada." They all
had experience with Ada in many contexts, but also experience in areas where Ada
wasn't used.
Frank Belz (TRW): many situations don't require a lot of programming or some
other language fits better. Many business applications don't use programming
languages in the third generation sense. Other situations related to market
perception and time to delivery. Ada is perceived as "not a research language."
University graduates hadn't been taught Ada. He thinks that GNAT and other
educational initiatives will help.
Other notes: all Ada vendors have programs to encourage the use of their
compilers in universities; Intermetrics and Thomson have teamed to produce an
academic oriented compiler to be sold in bookstores.
Mark Scott Johnson (Sun): it's not C. He also talked about the JAVA project.
JAVA shares many similar design goals with Ada. Didn't use C++ because it was
not small, not reliable, not safe. He also mentioned the efforts in the Ada
community (particularly Tucker Taft and Bob Munk) to compile Ada95 to JAVA byte
codes. He thought that JAVA timing (by luck) was right to generate so much
interest. He also expressed some of the growing industry sentiment against C++.
Sy Wong (Ada consultant): compared TRI-Ada'95 to a cocktail party on the deck of
the Titanic and the Ada community in general as being held captive in a DoD zoo
wondering why animals outside weren't interested in coming in to get free meals.
He thought Ada95 was a step backward in the simplicity of programming and
language size dimension. He talked about large production volume microprocessors
used in commercial control applications (like washing machines). Sy had worked
with hardware description languages in the 80's, in particular VHDL (which is
"Ada in disguise").
Bob Mathis (ARA) offered some responding comments about getting people to
reconsider Ada because of changed circumstances and improved features in Ada95.
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: John Barnes Plenary Talk
Date: 11 Nov 95 01:02:10 EST
The most entertaining talk so far was given this morning (Thursday) by John
Barnes. John was a member of both the Ada83 and Ada95 design teams and author of
the best selling Ada textbook. John traced the history of programming languages
with some conclusions about good features of Ada and weaknesses of some other
languages.
A few notable quotes and phrases: "Don't confuse action with progress." "Know
the difference between professionally engineered software and personal systems."
Information hiding should hide irrelevant details (not relevant ones). Ada
provides freedom from errors; this is more important than freedom to do anything
you want in a program. Multiple inheritance is the spaghetti of object-oriented
inheritance. Four good features of Ada95 -- object-oriented, protected types,
child libraries, and flexible access.
John summarized his opinion of different programming languages: Smalltalk is
flexible and reliable, C++ is flexible and efficient, Ada83 reliable and
efficient, but best of all Ada95 has all three (flexible, reliable, and
efficient).
So as not to have the same problems with tea as the day before, this time the
organizers arranged in advance for a martini (in fact more than one and later
revealed to be just stage props). There were also some interesting excursions
into geometric proofs.
From: J.M.KAMRAD.II@cdev.com (j.m.kamrad.ii)
Subject: Re: Microsoft and Ada
Date: Sun, 12 Nov 1995 13:27:03 -0500
>Chuck Engle announced at the closing plenary, that he had received official
>notice from Microsoft that they have accepted our Ada95-Win32 bindings and will
What is "our Ada95-Win32 binding"?
>be distributing them from their server. It is not a Microsoft product, but it
>is
>the first non-Microsoft language product they have endorsed.
------------------------------------
Mike Kamrad
Computing Devices International kamrad@cdev.com
M/S BLC W2J 1.612.921.6908
8800 Queen Avenue South FAX: 1.612.921.6552
Bloomington MN 55431
From: J.M.KAMRAD.II@cdev.com (j.m.kamrad.ii)
Subject: Re: Dave Emery Comments
Date: Sun, 12 Nov 1995 13:28:35 -0500
>Dave Emery had the general comment that the papers at the conference were very
>good overall (just a little plug for ACM, the conference proceedings are
>purchasable separately).
Where do we purchase the proceedings??
GNAT has triggered lots of exciting work. EPFL (Swiss
>Fed. Inst. of Tech.) continues to lead in object-oriented programming insights.
>More than 50% of "Best Paper" nominees were European.
>
>Arne Carlsson (Saab Ericcson Space) described a satellite system using
>RAD-hardened microprocessor with Ada-specific instruction set. He also asks why
>aren't there more Ada processors? We couldn't get performance under space,
>size,
>power, RAD, etc. constraints without hardware support for the language (for any
>language)
>
>p.s. Emery denies that his finger injury is due to strong typing or punching C
>programmers...
------------------------------------
Mike Kamrad
Computing Devices International kamrad@cdev.com
M/S BLC W2J 1.612.921.6908
8800 Queen Avenue South FAX: 1.612.921.6552
Bloomington MN 55431
From: Rush Kester
Subject: Re: Microsoft and Ada
Date: Sun, 12 Nov 1995 19:15:52 -0500 (EST)
On Fri, 10 Nov 1995, Robert F. Mathis wrote:
> Chuck Engle announced at the closing plenary, that he had received official
> notice from Microsoft that they have accepted our Ada95-Win32 bindings and will
> be distributing them from their server. It is not a Microsoft product, but it is
> the first non-Microsoft language product they have endorsed.
I see this as a very positive thing for Ada for two reasons. First,
having Ada code available from a very popular server is good exposure.
Second, this means one less "excuse" for not using the language.
Congratulations to all who helped make this happen.
Rush Kester
Team-Ada
From: Rush Kester
Subject: Re: Dave Emery Comments
Date: Sun, 12 Nov 1995 19:29:56 -0500 (EST)
On Fri, 10 Nov 1995, Robert F. Mathis wrote:
> Regarding Dave Emery's comments on the papers at the conference
> More than 50% of "Best Paper" nominees were European.
It seems the Europeans are less hung up on Ada's DoD origins or are
subject to less marketing hype from major software vendors like Microsoft,
Borland, etc. who are pushing C/C++. Or am I missing something happening
in Europe that would account for this? Perhaps something we should try
to emulate in the U.S.
Anyone else care to speculate?
Rush Kester
Team-Ada
From: Rush Kester
Subject: Re: Dave Emery Comments
Date: Sun, 12 Nov 1995 19:47:19 -0500 (EST)
On Sun, 12 Nov 1995, j.m.kamrad.ii wrote:
> Where do we purchase the proceedings??
Call the ACM at (800) 342-6626 and request Tri-Ada'95 proceedings.
Make sure you the year clear, I requested "this year's Tri-Ada
proceedings" and got 94 by mistake. The cost will probably be $36-50.
From: Cheryl Marquis
Subject: Re: TRI-Ada'96
Date: Mon, 13 Nov 1995 08:32:58 -0500
> TRI-Ada'96 will be held in Philadelphia, PA, December 1-6, 1996. Philadelphia is
> emphasizing 1996 as the "Year of the Computer" in honor of the 50th anniversary
> of the ENIAC at the University of Pennsylvania.
>
At the new Pennsylvania Convention Center?
Great I can visit attend the conference _and_ visit my family! ;-D Just one
question - Philadelphia in December? Brrr!
Cheryl
From: emery@grebyn.com (David Emery)
Subject: IEEE 990 (Ada as PDL)
Date: Mon, 13 Nov 95 18:40 PST
Well, my finger is a bit better so I can type a bit more... A small group
met at a BOF session at Tri-Ada to discuss the fate of IEEE Std 990. This
is the "Recommended Practice for the use of Ada as a PDL." The meeting
was chaired by Mark Gerhardt. The original plan was for Mark and I to
"debate" the issue, but instead we had a less formal meeting, so we could
come to consensus in time to attend the Ada and Java/WWW BOF later that
evening...
The topic is the status (fate) of IEEE 990, which is up for reaffirmation.
(IEEE Standards must be periodically reaffirmed, or else they die or suffer
some similar fate.)
The group decided that:
1. IEEE 990, as currently written, is not worth renewing. The
document (4 pages of 'normative text') has had little or no positive
effect on the 'business'. The group unaminiously voted to 'trash'
(Recycle, for you unfortunate Win95 users :-) the current standard.
2. There was interest in doing a 'design language/design notation'
standard. One suggestion was to 'standardize' the Booch notation, or
the forthcoming "grand unified OO notation" from Rational. At least one
opinion was expressed for a design notation that was language independent.
Therefore, the group will attempt to form a SIGAda WG to come up with
a general direction (if any) for a design notation standard, and once
this direction is reached, start work on an IEEE standard. In the interim,
the group wanted to 'gain jurisdiction' on the reaffirmation of the
IEEE 990 standard, to prevent some (less well informed) group from
trying to reaffirm the standard. It is an open issue, if the group
decides on a design notation approach, to either re-use the IEEE 990
activity (producing a complete revision), or to open another IEEE project.
dave
Subject: Re: IEEE 990 (Ada as PDL)
From: Jim Moore
Date: Tue, 14 Nov 1995 08:29:54 -0500
It's not quite correct to say that IEEE 990 (Ada as PDL) is
"up for reaffirmation". Reaffirmation is not required until 1997.
The reason that I asked for an expression of interest from SIGAda
at this particular time is because Ada 95 (apparently) obsolesces
the existing PDL, which, of course, is based on Ada 83.
I take Dave Emery's report as an expression of interest, albeit
a cautious one.
Dave mentions "gaining jurisdiction" on the reaffirmation of the
990 standard. That is not a term that I would use, but the folks who
would decide on the reaffirmation of 990 would be a balloting group
formed principally from the balloting pool that already exists for
the body of standards maintained by the IEEE Software Engineering Standards
Committee. (This balloting pool is an open group; anyone can join.)
I would welcome the formation of a SIGAda Working Group to provide advice
on the future direction of IEEE 990. Of course, the direct way to
implement such advice is to form an IEEE working group to draft a
replacement for 990.
Anyone who is interested in joining (or leading) such a working group
is invited to contact me. Rather than replying to the entire list
(risking annoying all of its readers), you should probably reply directly
to me at "moorej@acm.org".
Regards, Jim Moore
Member of Management Board, IEEE Software Engineering Standards Committee
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Tue, 14 Nov 95 23:33:02 -0500
It is quite practical to create ports of GNAT to a wide variety of
micro-controllers.
From: AdaWorks
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 09:34:25 -0800 (PST)
On Tue, 14 Nov 1995, Robert Dewar wrote:
> It is quite practical to create ports of GNAT to a wide variety of
> micro-controllers.
Robert,
I have not the slightest doubt that GNAT is portable to a wide variety
of micro-controllers. I am not sure of the 8051 since it is a very
odd design with limited stack management and limited memory management
capabilities among other things.
I can slice a tomato with a double-bitted axe, but I would probably
choose a small knife instead given my awareness of the tensile strength
of the tomato skin, that certain look I want it to have in my salad, and
the greater precision afforded me by the smaller implement.
I am certainly not opposed to GNAT. I am definitely impressed by what
you and your colleagues at NYU have accomplished, and congratulate you
on that. My question is really more concerned with the propriety of
any Ada compiler for the 8051 given the peculiarities of that processor's
architecture. It is a question, not an answer. If the answer, prepared
by someone intimately acquainted with both the 8051 and the new Ada
standard, is "Yes," I will be delighted.
My question simply reflects my skepticism regarding the potential for
inappropriate use of otherwise excellent tools.
Richard Riehle
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 95 14:09:34 -0500
if it practical to have a C compiler for the 8051, then it is practical
to have a GNAT port to it, it is that simple!
From: AdaWorks
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 12:57:42 -0800 (PST)
On Wed, 15 Nov 1995, Robert Dewar wrote:
> if it practical to have a C compiler for the 8051, then it is practical
> to have a GNAT port to it, it is that simple!
Robert,
Interesting phrasing. I particularly like your choice of the word,
"practical." There are those in the 8051 cult who would sieze upon
this word as justification to eschew the use of any high-level language
such as Ada or a universal assembler such as C.
Also, I would hesitate to go so far as to say that anything that is
practical in C is also practical in Ada. Even as a card-carrying Ada
advocate this seems a bit of a jump. Had you said C++ rather than C,
it would be easier for me to concur.
C, especially traditional K&R C, is fundamentally a universal assembler.
Stepanov calls it a "virtual machine." It is at a level of abstraction
that maps directly to the hardware. Ada is at a different level of
abstraction. For most environments this does not matter. For the
8051, I suspect that it does.
This is a function of the architecture of this odd little processor.
Perhaps some brave soul in the Ada community would like to prove that
the 8051 does indeed accomodate the demands of Ada by doing the port of
GNAT or writing a compiler.
If this can be done, and if it is done, I believe there would be a market
for it. That is, there would be a market it the compiler can compete with
the demands for efficiency expected of the typical 8051 application. I
continue to have my doubts, but will delight in being shown to be
dead wrong. In fact, I hope I am wrong.
My deal. If someone can do a solid port to the 8051, I will persuade the
editor of embedded systems programming magazine to let me do another
piece on Ada well before the current schedule of Tri-Ada 96. It will
certainly surprise the ESP community, much of which is focused on the
8051.
Richard Riehle
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 95 17:32:29 -0500
" Also, I would hesitate to go so far as to say that anything that is
practical in C is also practical in Ada. Even as a card-carrying Ada
advocate this seems a bit of a jump. Had you said C++ rather than C,"
This is a technically bizarre claim. The machine model of Ada at the
implementation level is VERY close to C, the high level semantics of
Ada have almost nothing to do with code generation (concepts like
private types, discriminants, packages etc have absolutely no code
generation consequences).
Can you at least give a hint of an example where a problem exists?
I especially find your idea that C++ is somehow easier even more
bizarre. What aspects do you consider to be substantially different
between Ada and C++ with respect to code generation.
I am really at a COMPLETE loss to understand what you are getting at
here. I can understand someone non-technical making such statements
based on some vague appreciation of what the implication of the three
languages is, but you must hjave some technical basis for your comment
that completely escapes me.
In terms of GCC, if you have a C compiler, you automatically have an
Ada compiler (admitedly without tasking, the tasking takes extra support)
and a C++ compiler. So what I said stands absolutely for the GCC
technology.
From: kst@thomsoft.com (Keith Thompson)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 14:53:44 PST
> In terms of GCC, if you have a C compiler, you automatically have an
> Ada compiler (admitedly without tasking, the tasking takes extra support)
> and a C++ compiler. So what I said stands absolutely for the GCC
> technology.
What about fitting the Ada runtime into a small memory space? Tasking
isn't the only feature of Ada that requires extra runtime support.
I believe that a sufficiently restricted Ada could be implemented for
any target that supports C. The question is, what does "sufficiently
restricted" mean? Is there reason to believe that it means dropping
tasking and only tasking?
--
Keith Thompson (The_Other_Keith) kst@thomsoft.com (kst@alsys.com still works)
TeleSoft^H^H^H^H^H^H^H^H Alsys^H^H^H^H^H Thomson Software Products
10251 Vista Sorrento Parkway, Suite 300, San Diego, CA, USA, 92121-2718
Because I'm weird enough, I'm sick enough, and doggone it, people fear me!
From: dweller@dfw.net (David Weller)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 16:54:40 -0600 (CST)
> In terms of GCC, if you have a C compiler, you automatically have an
> Ada compiler (admitedly without tasking, the tasking takes extra support)
> and a C++ compiler. So what I said stands absolutely for the GCC
> technology.
>
Robert,
I think Richard's point was whether one could reasonably expect to
"fit" GNAT onto an 8-bit CPU with a limited architecture. Perhaps
this is a question better left for TLD Systems, who intends to
specialize in creating GNAT-based cross compilers for embedded
processors.
In general, I share Richard's trepidation (big word of the day :-).
I think the effort to port GNAT to such a target would be non-trivial.
While it's very true that you get "C equivalent" performance code with
GNAT, that's a far leap from claiming you can fit all Ada features
neatly into an itty-bitty CPU. Of course, you don't have to use them
all, but I _think_ Richard's point was that, once you trimmed Ada down
to what you _should_ use, you'd wind up with something close to C.
Of course, all of this simply points out (screams?) that a "microset"
of Ada would probably help a lot in places where you don't want so
much "Stuff". It would mean deleting things like generics, tasks, and
some of the more esoteric features, while permitting us to keep
helpful things like namespaces, subprogram pointers, and Ada's strong
typing.
Ah, wishful thinking, I suppose.
pragma Itty_Bitty_Living_Space; :-)
From: "Robert I. Eachus"
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 18:16:51 -0500
(I hope my trimming helped some of you recieve only one copy...)
There are several Ada compilers which have very small linked-in
run-time libraries. If I were to do a GNAT port to something with a
16-bit address space, I would completely review all the run-time to
see what could be shrunk, in many cases at the cost of slower
performance or limited functionality. (For example, only supporting a
fixed number of tasks, and replacing a lot of the tasking structures
built "on the fly" with a table of task descriptors, sized at
link-time.)
Similarly, a lot of the run-time should be distributed to multiple
object modules so that only the necessary parts of, say, Text_IO get
linked in. This is what I meant by my "tenth on the list" comment,
there is a lot of grunt work to do if you want to target a small
machine. (Also you will need libraries to do multiprecsion integer
arithmetic, etc. These exist, but you want to integrate with the code
generator to use threaded interpreted code style calls, so code using
32-bit integer arithmetic will be very compact, if somewhat slow.)
If someone is willing to put forth this effort, and Terry may be,
then the limit on what you can do in Ada on an 8-bit chip will be very
similar to what you can do in C. (But the source code will be a lot
nicer. ;-)
Robert I. Eachus
with Standard_Disclaimer;
use Standard_Disclaimer;
function Message (Text: in Clever_Ideas) return Better_Ideas is...
From: AdaWorks
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 15:57:31 -0800 (PST)
On Wed, 15 Nov 1995, Robert Dewar wrote:
> This is a technically bizarre claim. The machine model of Ada at the
> implementation level is VERY close to C, the high level semantics of
> Ada have almost nothing to do with code generation (concepts like
> private types, discriminants, packages etc have absolutely no code
> generation consequences).
>
> Can you at least give a hint of an example where a problem exists?
I guess I have to root around through old piles of technical material
for some of my 8051 stuff. However, I'm not saying this cannot be
done, but it seems to me that Ada is not going to be a simple port
given the requirement for a run-time environment to do type
checking, tasking, generics, etc. A pared down version? I don't
know.
> I especially find your idea that C++ is somehow easier even more
> bizarre.
Sorry to have misled you on my intention here. I did not mean to
imply that C++ is easier. In fact, it is probably more messy.
> I am really at a COMPLETE loss to understand what you are getting at
> here. I can understand someone non-technical making such statements
> based on some vague appreciation of what the implication of the three
> languages is, but you must hjave some technical basis for your comment
> that completely escapes me.
I will try to have a few more words on this when I find my old 8051
stuff. However, it will surprise me if the person who finally makes
the Ada compiler for the 8051 comes close to the expectations of
those who are still programming this little monster.
> In terms of GCC, if you have a C compiler, you automatically have an
> Ada compiler (admitedly without tasking, the tasking takes extra support)
> and a C++ compiler. So what I said stands absolutely for the GCC
> technology.
I hope you are right. However, I reserve the right to be a little
skeptical with respect to the 8051 until some actually completes an
acceptable Ada compiler for it. If this actually happens, it will
serve notice on everyone who ever wanted a waiver to use something
other than Ada will be without any foundation for that waiver.
I would greatly enjoy the opportunity to report to my JOOP readers
and especially my Embedded Systems Programming magazine readers, that
there has been a successful version of Ada targeted to the 8051.
Tucker indicated in his presentation on programming languages that
language does make a difference. I believe that microprocessor
architecture also makes a difference. For example, I wonder how
easy it would be to port GNAT to a transputer? How easy would it be
to port GNAT to a Massively Parallel Architecture?
Perhaps I do not have enough knowledge of the internal model of GNAT,
but I wonder if it is really a universal compiler-compiler. Or are
we to believe that it is unlimited in its power to target any
architecture?
Robert, I do not intend this as sarcasm. Probably my ignorance of the
full power of GCC is showing. But I do know that some of these little
embedded processors are characterized more by the peculiarities of their
architecture than by their commonality.
Thanks engaging in this discussion. It is helping me to understand
some of the issues relative to GNAT.
Richard
From: AdaWorks
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 16:20:34 -0800 (PST)
On Wed, 15 Nov 1995, David Weller wrote:
> Robert,
>
> I think Richard's point was whether one could reasonably expect to
> "fit" GNAT onto an 8-bit CPU with a limited architecture. Perhaps
> this is a question better left for TLD Systems, who intends to
> specialize in creating GNAT-based cross compilers for embedded
> processors.
I rather doubt that even TLD will tackle the 8051. They have done
some excellent work on the 1750A which Intelsat uses on its
communications satellites. These are programmed in Ada and they
are a 16-bit architecture. The 1750A also has an odd architecture
and early compilers were restricted to a small memory model. Current
compilers support the 1750A MMU, but not everyone uses it.
> In general, I share Richard's trepidation (big word of the day :-).
> I think the effort to port GNAT to such a target would be non-trivial.
A fair characterization of my view. -- RR
> While it's very true that you get "C equivalent" performance code with
> GNAT, that's a far leap from claiming you can fit all Ada features
> neatly into an itty-bitty CPU. Of course, you don't have to use them
> all, but I _think_ Richard's point was that, once you trimmed Ada down
> to what you _should_ use, you'd wind up with something close to C.
The fact is that many 8051 programmers don't even like the result they
get from available C compilers.
> Of course, all of this simply points out (screams?) that a "microset"
> of Ada would probably help a lot in places where you don't want so
> much "Stuff". It would mean deleting things like generics, tasks, and
> some of the more esoteric features, while permitting us to keep
> helpful things like namespaces, subprogram pointers, and Ada's strong
> typing.
Strong typing at the compiler level. I'm not certain how one would
implement the RTE for the 8051. How big is the smallest executable
for a program compiled by GNAT for a source code program of , say,
1,000 SLOC?
I really hope someone will do this port and establish once-and-for-all
that I am totally wrong on this one. If you start this project, be
prepared for an interesting, and lengthy, experience. You should be
able to sell the resulting compiler if it can compete with the usual
way of programming the 8051.
Richard Riehle
From: AdaWorks
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 16:49:16 -0800 (PST)
On Wed, 15 Nov 1995, Robert I. Eachus wrote:
> see what could be shrunk, in many cases at the cost of slower
> performance or limited functionality. (For example, only supporting a
> fixed number of tasks, and replacing a lot of the tasking structures
> built "on the fly" with a table of task descriptors, sized at
> link-time.)
INteresting notion.
> Similarly, a lot of the run-time should be distributed to multiple
> object modules so that only the necessary parts of, say, Text_IO get
> linked in.
In fact, none of Text_IO for the typical 8051 application.
> there is a lot of grunt work to do if you want to target a small
> machine.
You betcha. And forget about re-entrant code, recursion, etc.
> (Also you will need libraries to do multiprecsion integer
> arithmetic, etc. These exist, but you want to integrate with the code
> generator to use threaded interpreted code style calls, so code using
> 32-bit integer arithmetic will be very compact, if somewhat slow.)
There will be no 32-bit integers, no floating point in the usual sense,
and no secondary memory. Generics, if supported at all, will have to be
based on some kind of table model rather than in-line code, and there
will be little point in defining any types other than Integer as defined
in package standard.
What remains of Ada will be the syntax associated with control structures
and some predefined types. Exception handling depends on a run-time
environment so that will be extraneous. Pragma Suppress(All_Checks) will
be standard practice and probably implied by a,
pragma Restrictions(I8051);
> If someone is willing to put forth this effort, and Terry may be,
I would be surprised to see Terry (if you mean Terry Dunbar) take the
time to invest in this kind of technology.
> then the limit on what you can do in Ada on an 8-bit chip will be very
> similar to what you can do in C. (But the source code will be a lot
> nicer. ;-)
I agree that the code would be nicer. However, if one finds that many
of the nicer capabilities of Ada disappear, what do we gain?
It is very exciting that someone out there might just be inspired to
pursue this as a project. Perhaps this would make a good Master's
thesis for someone.
Richard Riehle
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 95 20:08:41 -0500
"In general, I share Richard's trepidation (big word of the day :-).
I think the effort to port GNAT to such a target would be non-trivial.
While it's very true that you get "C equivalent" performance code with
GNAT, that's a far leap from claiming you can fit all Ada features
neatly into an itty-bitty CPU. Of course, you don't have to use them
all, but I _think_ Richard's point was that, once you trimmed Ada down
to what you _should_ use, you'd wind up with something close to C."
Who ever said that you could "fit all of Ada features". Obviously for
example one feature of Ada is million element floating-point arrays.
They won't fit!
On the other hand, it is quite false that "you'd wind up with something
close to C". This is nonsense. Most of the features of Ada (here is
a partial list: packages, child packages, private types and information
hiding, discriminants, named parameters, overloading, etc. etc. etc.)
have absolutely ZERO effect on size of generated code, and are fully
usable in the tiniest environment.
"Of course, all of this simply points out (screams?) that a "microset"
of Ada would probably help a lot in places where you don't want so
much "Stuff". It would mean deleting things like generics, tasks, and
some of the more esoteric features, while permitting us to keep
helpful things like namespaces, subprogram pointers, and Ada's strong
typing.
"
No it doesn't point out this, let alone scream it. A microset would
not help the compiler one little bit. The only point is to restrict
what a user can use, and I see no reason to predefine that. For
examle, it is a TERRIBLE idea to remove generics. Generics provide
a very powerful parametrization method that is usable in the tiniest
of environments.
I really hate to see Ada advocates spreading this kind of misinformation.
Maybe I am missing something ... what possible benefits could the definition
of such a microset have?
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 95 20:16:28 -0500
" I hope you are right. However, I reserve the right to be a little
skeptical with respect to the 8051 until some actually completes an
acceptable Ada compiler for it. If this actually happens, it will
serve notice on everyone who ever wanted a waiver to use something
other than Ada will be without any foundation for that waiver."
But in fact the major problems that the 8051 poses have nothing to do with
Ada specifically. That is why I am still so puzzled. The major problem
has to do with stack maintenance, which equally applies to C.
" Perhaps I do not have enough knowledge of the internal model of GNAT,
but I wonder if it is really a universal compiler-compiler. Or are
we to believe that it is unlimited in its power to target any
architecture?
Robert, I do not intend this as sarcasm. Probably my ignorance of the
full power of GCC is showing. But I do know that some of these little
embedded processors are characterized more by the peculiarities of their
architecture than by their commonality.
"
Nobody said that GCC could target any architecture, although we don't have
any examples of it failing in this task (rememebr that Motorola uses
GCC for most of its small microcontrollers).
What I said is quite simple, and 100% accurate, if there is a GCC port,
then you can easily get a GNAT port without tasking. This is not mysterious,
after all we know perfectly well that Ada can be compiled into C source
code if necessary. Of course you won't be able to run arbitrary Ada
programs, but then you won't be able to run arbitrary C programs as
well.
Please cough up ONE technical example where Ada has a harder time than
C in mapping to a microprocessor in this class. Without at least one
example, your argument is VERY unconvincing, and I believe that the
point you are making is badly misinformed.
Going back to the 8051, it is in fact quite difficult to implenment full
C semantics on this processor as you know. It maybe that for this processor
the predicate "if C can be successfully implemented" is simply false,
in which case the conclusion that Ada can be successfully implemented
is equally false (but no more false).
In other words, suppose we find that it is impossible to implement recursion
practically on the 8051, which may nbe the case.
Well a C compiler would have the limitation that a program could not make
any recursive calls.
An Ada 95 compiler would have EXACTLY the same restriction (only though
it would be a little neater, since one would document and at least to
a limited extent enforce this restriction at compile time by using
pragma Restrictions (No_Recursion)
as a configuration pragma for the partition.
I really think you are selling Ada 95 badly short here.
I also wonder if you are aware of the wide variety of Microcontrollers
for which GCC ports exist?
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 95 20:25:07 -0500
"> In general, I share Richard's trepidation (big word of the day :-).
> I think the effort to port GNAT to such a target would be non-trivial.
A fair characterization of my view. -- RR"
If there is a GCC port, then it is very easy to create the GNAT port. If
there is no GCC port, then one has to be done first. If there are existing
C compilers for the 8051, then it is certainly practical to do at least
as well as these C compilers using GCC.
" Strong typing at the compiler level. I'm not certain how one would
implement the RTE for the 8051. How big is the smallest executable
for a program compiled by GNAT for a source code program of , say,
1,000 SLOC?"
The answer to that question is that it is no larger than the same
answer for a 1,000 SLOC C program. The minimum size of an Ada program
is very tiny (right now the overhead is about 500 bytes, but this
could be reduced greatly by taking advantage of pragma Restrictions).
Now of course if you use Text_IO, you need the Text_IO stuff, but that's
equally true of printf. GNAT, like any C compiler, only loads runtime
routines that are actually used. Let's look at the basic required part
of the runtime in GNAT:
It consists of the following units:
with System.Tasking_Soft_Links;
about 50 bytes, could be eliminated if there was no tasking
with System.Task_Specific_Data;
about 50 bytes, could be eliminated if there was no tasking
with System;
zero bytes
with System.Storage_Elements;
about 80 bytes, could be reduced with a bit of work
with System.Secondary_Stack
about 300 bytes, can be eliminated if there is no dynamic allocation
with Unchecked_Conversion;
-- Referenced from System.Secondary_Stack and System.Task_Specific_Data
with Unchecked_Deallocation;
these are both zero bytes
Now, as I say, if you use features that require runtime routines, e.g.
you raise an exception or compute a random number, then you need the
corresponding runtime code, but that's true of C too.
I fail to see even one small technical suggestion of a fundamental
difference between C and Ada here. Ada is a low level language in terms
of its machine model, with minimal standard runtime requirements.
I think you are biased because of your experience with other Ada
technologies, but what you are saying has nothing to do with the
language, and in the case of GNAT, with the implementatoiun either.
From: AdaWorks
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 19:01:18 -0800 (PST)
On Wed, 15 Nov 1995, Robert Dewar wrote:
DW> "> In general, I share Richard's trepidation (big word of the day :-).
DW> > I think the effort to port GNAT to such a target would be non-trivial.
RR> A fair characterization of my view. -- RR"
RD> If there is a GCC port, then it is very easy to create the GNAT port. If
RD> there is no GCC port, then one has to be done first. If there are
existing
RD> C compilers for the 8051, then it is certainly practical to do at least
RD> as well as these C compilers using GCC.
RR: I am not sure whether there is GCC port. There are C compilers
RR: but many 8051 programmers continue to prefer assembler.
RD> The answer to that question is that it is no larger than the same
RD> answer for a 1,000 SLOC C program. The minimum size of an Ada program
RD> is very tiny (right now the overhead is about 500 bytes, but this
RD> could be reduced greatly by taking advantage of pragma Restrictions).
RR: OK. Five hundred bytes is certainly small enough. Perhaps you are
RR: correct in a later statement in this message that I am still stuck
RR: in the typical code sizes for earlier Ada technology.
> Now of course if you use Text_IO, you need the Text_IO stuff, but that's
> equally true of printf.
It is rare, in an 8051 application, to require a feature such as Text_IO.
One could probably eliminate it altogether for most applications. Since
the typical application is single-threaded, tasking would be unnecessary.
Also, floating point would be unnecessary for many applications since
the kinds of devices being controlled are often discrete. There are, of
course, some applications that might require some form of decimal
fraction, but usually much more primitive than a full-blown floating
point model. There are tons of stuff that can be optimized away for
most 8051 applications.
> GNAT, like any C compiler, only loads runtime
> routines that are actually used. Let's look at the basic required part
> of the runtime in GNAT:
>
> It consists of the following units:
>
> with System.Tasking_Soft_Links;
>
> about 50 bytes, could be eliminated if there was no tasking
RR: Then this would be eliminated under nearly all circumstances.
> with System.Task_Specific_Data;
>
> about 50 bytes, could be eliminated if there was no tasking
RR: Same as above
>
> with System;
>
> zero bytes
RR: Can't get much better than that.
>
> with System.Storage_Elements;
>
> about 80 bytes, could be reduced with a bit of work
>
> with System.Secondary_Stack
>
> about 300 bytes, can be eliminated if there is no dynamic allocation
RR: Dynamic allocation is rare on 8051. This is even rare in the Ada
RR: applications that use 16-bit processors, in my modest experience.
> with Unchecked_Conversion;
> -- Referenced from System.Secondary_Stack and System.Task_Specific_Data
RR: Doubt whether this would be applicable.
>
> with Unchecked_Deallocation;
RR: Probably never applicable.
> Now, as I say, if you use features that require runtime routines, e.g.
> you raise an exception or compute a random number, then you need the
> corresponding runtime code, but that's true of C too.
>
> I fail to see even one small technical suggestion of a fundamental
> difference between C and Ada here. Ada is a low level language in terms
> of its machine model, with minimal standard runtime requirements.
>
> I think you are biased because of your experience with other Ada
> technologies, but what you are saying has nothing to do with the
> language, and in the case of GNAT, with the implementatoiun either.
RR: I think I fell into a trap of my own design by making the comparison
RR: between Ada and C. You know I am no fan of C. The original point
RR: was my sense of the difficulty of taking advantage of the benefits
RR: of Ada on the 8051. You have indicated a few such as recursion and
RR: stack management and memory size. If GNAT is as efficient in its
RR: use of memory as you suggest, perhaps there is no problem. However,
RR: I am looking forward to seeing the actual Ada compiler for the
RR: the 8051 when it is ready. As I stated earlier, I will be pleased
RR: to discover I have been wrong when that compiler is a reality.
RR: This discussion has ensued without my having a chance to locate my
RR: old 8051 stuff. If I can find it (particularly to review the
RR: register model) I shall reply to Robert individually without
RR: bothering everyone else out there.
RR: Meanwhile, who is going to do the GNAT/8051 port?
Richard
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 95 22:22:56 -0500
" It is rare, in an 8051 application, to require a feature such as Text_IO.
One could probably eliminate it altogether for most applications. Since
the typical application is single-threaded, tasking would be unnecessary.
Also, floating point would be unnecessary for many applications since
the kinds of devices being controlled are often discrete. There are, of
course, some applications that might require some form of decimal
fraction, but usually much more primitive than a full-blown floating
point model. There are tons of stuff that can be optimized away for
most 8051 applications."
The point is that "optimized away for most 8051 applications" is not the
issue, you pay for this stuff only if you use it. Removing it from GNAT
would not help one iota in the task of generating an Ada compiler for the
8051.
"Who is going to do the GNAT/8051 port?"
I doubt there is a market for such a port, so I doubt anyone will do it.
If someone is interested in a GNAT port for this chip and willing to pay
for it, we (ACT) would be glad to investigate.
From: BSCrawford@aol.com
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 1995 22:33:33 -0500
In a message dated 95-11-15 David Weller writes:
>Of course, all of this simply points out (screams?) that a "microset"
>of Ada would probably help a lot in places where you don't want so
>much "Stuff". It would mean deleting things like generics, tasks, and
>some of the more esoteric features, while permitting us to keep
>helpful things like namespaces, subprogram pointers, and Ada's strong
>typing.
Isn't this suggestion more or less the same thing that Sy Wong has been
asking for? I hope this isn't a dumb question, but this discussion is
beginning to make me wonder whether there may be a significant
opportunity for Ada in the small system/small DSP area.
Bard Crawford
Stage Harbor Software
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Wed, 15 Nov 95 23:32:28 -0500
Sy Wong, like Dave Weller, somehow seems to think that defining an Ada
subset will help implementation on small machines. As long as you are
not trying to *host* the compiler on the small machine, this thought is
100% bogus. Limiting the language will not in any way help the compiler
generate code for small machines. If there are cases where limiting
the *programs* (note the difference) helps, this can be achieved by
use of pragma Restrictions.
From: Rush Kester
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Thu, 16 Nov 1995 01:03:28 -0500 (EST)
Continuing a thread started on TriAda95 on Team-Ada for those who
missed some postings. If you get multiple copies, temporarily, my
apologies, but now that TriAda is over, Team-Ada is a more appropriate
forum. If you reply, please delete Tri-Ada from the list of addressees.
Perhaps the real question that this seems to be leading to is not
"Can a cross-compiler be built for a particular microprocessor/
microcontroller?",
but rather
"For a given application, could a prudent Ada developer use the GNAT
(or some other equally economical) cross-compiler to generate code that
will run as efficiently as if a C (or some other currently acceptable
language) cross-compiler would?"
The above question, only those a few in the Ada community familiar with
cross-compiler technology can address. Robert Dewar, I believe is one
qualified candidate, any one else qualified to respond?
A corralary (sp?) is then,
"Is there sufficient market for the given microprocessor/
microcontroller for more than one easily ported cross-compiler to
be built?"
This second question, only those in the microprocessor/microcontroller
market can effectively address.
Rush Kester
Team-Ada
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Mathis DRAFT Trip Report
Date: 16 Nov 95 08:34:16 EST
After the TRI-Ada'95 Conference, this is my attempt at a summary trip report.
Other people said they would also try to write summaries and post them to this
list. I had a good time reporting on the Conference and I hope that people got
something out of it. The best sessions were undoubtedly the ones I didn't
attend, so I appreciate others adding to my comments. -- Bob Mathis
- - - - - - - - - - - - - - - - - - - - - - - - - -
My own personal conclusion at the end of the conference was very positive. This
is much better than the ending feeling of some other recent Ada conferences.
Final registration was over 600 (more than expected). Terry Doran, Jim Moore,
and other members of the committee deserve our thanks and congratulations. Job
well done.
On the other hand, I think the current TRI-Ada model needs considerable work.
SIGAda is planning the next big conference for December 1-6, 1996, in
Philadelphia. Comments and suggestions are welcome. This list
(TriAda95@ocsystems.com) is a good place to discuss TRI-Ada'95, reactions to it,
and suggestions for next year. After this list goes away (November 30, 1995),
the ARA list (ara@ocsystems.com) is an appropriate place to discuss Ada
conferences.
Vendor Announcements
Intermetrics announced the first Ada95 validation two weeks ago. ACT and SGI
validated on the exhibit floor. OC Systems is 98 and 44/100s complete. R&R
Software was giving away some free copies of their Ada95 compiler for Windows NT
or Windows 95. Tartan announced their SHARK compiler. TLD will be going with
GNAT for Ada95. Thomson has teamed with Intermetrics for the production of a
student compiler and then professional level products. Rational continues to
offer a broad range of products and migrate them to Ada95. DEC is teaming with
Rational and ACT for the different Alpha operating systems. DDC-I announced
their expectation of validating by the end of 1995. GreenHills was demonstrating
their multi-lingual, multi-platform tools. Irvine Compiler is migrating their
existing compiler to Ada95.
Chuck Engle (head of the AJPO) announced at the closing plenary, that he had
received official notice from Microsoft that they have accepted our Ada95-Win32
bindings (developed by Intermetrics, tested by LabTec, accepted by Ada compiler
developers in the Windows95 and Windows NT environments) and will be
distributing them from the Microsoft server.
There were many other vendors. They were offering advanced specialized tools
and/or services. I hope that attendees found what they were looking for and I
hope that vendors will feel free to post to this list information about their
activities, products, and successes.
Plenaries
At the main opening session, Hal Hart (SIGAda Chairman) presented SIGAda awards
to Chuck Engle, Mark Gerhart, Rick Conn, Robert Dewar, Jean Ichbiah, and Tucker
Taft.
Silicon Graphics provided some advance displays for the opening plenary session
(Tuesday). It gave the session a big conference, high-tech feel. John Mashey
(Silicon Graphics) gave the opening keynote stressing the trends in technology
to larger and faster computing and how this would change programming and human
interaction. It was a very nice keynote presentation looking toward the future
and opening minds to new approaches. Software innovation will be essential for
bridging the gap between more capable hardware and pretty much the same wetware
(human beings). E-mail is a very low-bandwidth communication medium and really
inadequate to convey the details, or the content, or the feel of this
presentation. Sorry, you had to be there.
In the Wednesday plenary, Richard Stallman talked about the historical
background and philosophy of the Free Software Foundation. "Free" means freely
usable and redistributable under the "copy-left" arrangement, not necessarily
free of cost. GNU and GCC served as the underlying framework for GNAT. It was an
interesting, but rambling talk. GNAT has changed the way people think about the
availability of Ada compilers. It was very useful to have Stallman himself
describe his philosophy, which he thinks of as a moral position. Robert Dewar,
head of the GNAT project, has restated and explained this philosophy frequently
in various Ada forums including a local LA SIGAda talk the night before.
Stallman expressed a general approval for the changes made in Ada95. He would
have made overload resolution less dependent on the context and redefinition of
functions on tagged types require a more explicit declaration of intent by the
programmer, but he didn't consider these major. At the end of the talk, there
was a discussion of potential changes in copyright law which would have a
negative effect on free software and other intellectual property. He pointed to
an upcoming article in the January issue of Wired.
John Barnes gave the Thursday morning plenary address. John was a member of both
the Ada83 and Ada95 design teams and author of the best selling Ada textbook.
John traced the history of programming languages with some conclusions about
good features of Ada and weaknesses of some other languages. Ada provides
freedom from errors; this is more important than freedom to do anything you want
in a program. Multiple inheritance is the spaghetti of object-oriented
inheritance. Four good features of Ada95 -- object-oriented, protected types,
child libraries, and flexible access. John summarized his opinion of different
programming languages: Smalltalk is flexible and reliable, C++ is flexible and
efficient, Ada83 reliable and efficient, but best of all Ada95 has all three
(flexible, reliable, and efficient).
Peter Coffee, PCWeek, gave the Friday plenary address and talked about an Ada
program he had done to exercise the Pentium division bug about a year ago. He
also discussed the difficulty others had in translating his program into C. He
talked about requirements for high reliable microprocessor software used in
non-computer appliances. The fraction of software that goes into desktop
applications is very small. Peter has very positive feelings about the
usefulness of Ada and concerns about the popularity of C++; but Ada is being
mentioned less and less in magazine articles. He mentioned John McCormick's
positive experience using Ada in a project course. He encouraged Ada people to
be more active in publicizing their successes.
Paper Sessions
Steve Zeigler's presentation, describing the experience of Verdix (now part of
Rational) in developing and maintaining its code base--which is about equally
divided between Ada and C code--provides compelling evidence that belongs in the
arsenal of every Ada advocate. Steve has been collecting data for a number of
years and has analyzed it carefully to discover patterns. Here are some of the
most striking results he presented:
- Ada code required 4.59 fixes per thousand lines of code, compared with 9.21
for C.
- Ada code contained .096 customer-reported defects per thousand lines of code,
compared with .676 for C.
- Development costs for Ada code were $6.62 per noncomment/nonblank line,
compared with $10.52 for C.
There were some very good papers and presentations at this conference. I
attended a few sessions and reported on them informally. I really want to
encourage others (even the authors themselves) to post summaries.
The Best Paper award went to "DVM: An Object-Oriented Framework for Building
Large Distributed Ada Systems," by Christopher J. Thompson (Hughes Aircraft) and
Vincent Cellier (Hughes Canada). The Best Presentation award went to "Ada95 as
Implementation Language for Object-Oriented Designs," by Stephane Barbey (Swiss
Federal Institute of Technology).
SIGAda Working Groups
The Working Groups are an important part of SIGAda's activities. ASISWG
announced its new draft and distributed it on disk (it will be available on
their Web page in a couple of weeks). There was discussion about a licensing
arrangement to encourage sharing Ada code without losing commercial rights. JAVA
and Ada were discussed at a number of sessions. (JAVA does not stand for "Just
Another Version of Ada.") Some of the working groups and the executive committee
got bogged down in administrative and procedural details. Information about
working group activities are available on the SIGAda Web page
http://www.acm.org/sigada/. Most of the working groups are reorienting their
work toward the different environment of Ada95.
Other Information
Patrick McDermott (a technical recruiter from Phoenix, AZ) (qrp@aol.com) said he
had over 450 job listings for Ada programmers. These are sales openings,
permanent programming openings, consulting, and so forth. The word needs to get
out that there are Ada jobs out there.
The Ada Resource Association is developing a new poster, Ferruccio's "The Vision
of Ada." Readers of these notes can request one by sending e-mail with the
subject "ARA Poster" to 73313.2671@compuserve.com (that's Bob Mathis the
Executive Director of the ARA). Please send the usual mailing information plus
phone, fax, and other relevant contact information. The ARA also has some
remaining copies of last year's poster showing a surfer. You can also request
one of these via e-mail.
Paul Whittington offered AdaSAGE CD ROMs "to all who E-Mail
me with a request to do so. Please include your complete land address. I will
put you on a list in as received order. We will mail to list members in order
until we run out of CDs so get your order in ASAP." Or you can download any or
all of the CD ROM from ftp://sageftp.inel.gov/pub/sage/cdrom002.
Ada Policy Summary
Chuck Engle was passing out some small cards describing Ada Policy. I've copied
the contents here because it's a short summary of what's important. The AdaIC
(800-232-4211 or 703-681-2466) has copies if you want one.
Why Ada?
Why is Ada appropriate?
Support for large, complex systems
Interoperability and maintainability
Software Engineering
Modifiable, reliable, portable, easily integrated, etc.
Economics
DoD core competency, lower lifecycle costs
International standard (ANSI, ISO, FIPS)
Only internationally-standardized object-oriented language
Only language with required validation
Promotes reuse, portability
Not locked into proprietary vendor
Most companies settle on a standard, why not DoD
Metrics
60-80% of software costs are in maintenance
Ada best in FAA and SEI scores (capability, cost, risk, etc.)
Ada leads in MITRE reliability and maintainability comparisons
Ada Policies
DODD 3405.1 Ada is the preferred common HOL.
Based on lifecycle cost, prefer use of : (1) COTS and advanced software
technology, when no government modification or maintenance during lifecycle; (2)
Ada; (3) DoD-approved standard HOL, if waiver granted.
Use Ada for all major upgrades (1/3 or more of lines total).
Army extensions: HQDA ltr 25-92-1, 25-95-1
Ada for all modifications of 1/3 or more of functional component.
SQL is approved for DBMSs.
4GLs permitted for prototypes, short-term, ad-hoc systems, non-Ada prototype
cannot be fielded.
Navy extensions: NAVINST5234.2A
Ada for modifications of 1/3 or more of computer software configuration item or
sub-system specification, within 5 years.
Waivers granted only on substantiation of economic analysis.
Air Force extensions: SAF/AQK Action memo
Distinguishes exceptions/waivers, gives details on each.
Exempts individual-use, unique, in-house applications.
SAF/AQK Info Memo Interprets term "cost effective" in Congressional Ada mandate
All three Services permit baselined ("project-validated") compilers - projects
can keep using same compiler throughout lifecycle (after validation certificate
expires).
Ada Information Clearinghouse
800/232-4211 or 703/681-2466
adainfo@sw-eng.falls-church.va.us
URL http://sw-eng.falls-church.va.us
Defense Information Systems Agency, Center for Software
Ada, The Language For a Complex World
-- -- -- -- -- -- -- -- End of Report -- -- -- -- -- -- --
From: "Mr. Michael Berman"
Subject: Cross posting: one word - don't!
Date: Thu, 16 Nov 1995 08:36:07 -0500
Please stop and think for a moment before cross posting messages to
this list and Team Ada. Then don't. The announcement that this list
exists went out to Team Ada and anyone and everyone that wants to
be on it already is. By cross posting to both lists, the net effect
is that everyone gets two of everything. It may not be coincidence that
during all of this 8051 thread (which belongs on comp.lang.ada in the
first place, but that issue is for another day) there have been two
public unsubscriptions from Team Ada, indicating that there were
probably several more successful ones. In other words, those of us who
cross post and the others of us who perpetuate it by blindly hitting the
reply key succeed only in scaring people _away_ from Team Ada, which is
the exact opposite of why Team Ada exists in the first place!
This message is posted only to triada95, but it's addressed to the Ada
Teamers as well. Sorry for the schoolteacher attitude, but it's in the
best interest of both lists and Ada herself.
(And yes, I know that this problem goes away in two weeks when this list
vaporizes. Consider this to be "scolding in advance" for when it occurs
on Brad's new AdaJava list!)
From: "Robert I. Eachus"
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Thu, 16 Nov 1995 10:01:28 -0500
Richard Riehle said:
> I really hope someone will do this port and establish once-and-for-all
> that I am totally wrong on this one. If you start this project, be
> prepared for an interesting, and lengthy, experience. You should be
> able to sell the resulting compiler if it can compete with the usual
> way of programming the 8051.
AFAIK the smallest computers that Ada has been targeted to
include the 128-kbyte Western Digital P-machines, the Russian PDP-11
clones, and the RRSoft compiler which ran on 512K 8088 based
PC-clones.
However, in all these cases the compiler was self-hosted!
So certainly an 8051 target is very reasonable, it just remains
to be seen if anyone wants to do it.
Robert I. Eachus
with Standard_Disclaimer;
use Standard_Disclaimer;
function Message (Text: in Clever_Ideas) return Better_Ideas is...
From: stt@dsd.camb.inmet.com (Tucker Taft)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Thu, 16 Nov 1995 10:04:34 -0500
I agree with Robert that Ada does not impose any particular
challenge over and above the challenge of building any high-level
language compiler for a brain-dead instruction set.
Certainly if you use certain Ada features like tasking, you
will get additional out-of-line code linked in, but there were
many Ada compilers built for 16-bit machines, and some of them
had extremely compact run-time systems.
As Robert points out, however, most of Ada's important features
involve little or no additional code at run-time. Things like
generics, private types, very strong type checking, etc., involve
no run-time overhead. Tasking and exception handling do both
involve out-of-line code, but not as much as you might think, and
with protected types, there is probably little need for using
rendezvous in a small application.
The real point is that some processors (DSPs, 4-bit microcontrollers, etc.)
are a pain to program in C, Ada, Modula-2, Eiffel, whatever.
Their instructions sets are just too feeble to support any
high-level language. However, as there is a robust market for
8051 C compilers, I presume the 8051 could be just as easily the
target for an Ada compiler. I agree with Robert that it is silly
to recommend C over Ada because somehow the instruction set implies
it is a better fit to C. The C and Ada run-time models are extremely
close. However, one could legitimately recommend C over Ada because the C
compiler already exists, and the Ada compiler doesn't. However, if the
C compiler that exists is GCC, then getting a version of GNAT is simple
enough to warrant investigation.
-Tucker Taft stt@inmet.com
Intermetrics, Inc.
From: westley@hercules.calspan.com (Terry J. Westley)
Subject: Re: Team-Ada mailing list
Date: Thu, 16 Nov 1995 11:32:58 -0500 (EST)
> Continuing a thread started on TriAda95 on Team-Ada for those who
> missed some postings. If you get multiple copies, temporarily, my
> apologies, but now that TriAda is over, Team-Ada is a more appropriate
> forum. If you reply, please delete Tri-Ada from the list of addressees.
Since some of my 8051 code flew in Desert Storm, I'm very interested
in this thread. Can you tell me how to subscribe to the Team-Ada
mailing list in case it re-appears over there instead of on comp.lang.ada
when triada95 goes away?
--
Terry J. Westley, Principal Computer Scientist
Calspan SRL, P.O. Box 400, Buffalo, NY 14225
westley@calspan.com http://worf-gw.calspan.com/~westley/
From: woodruff@tanana.llnl.gov (John Woodruff)
Subject: Seeking copy of best paper
Date: Thu, 16 Nov 1995 09:30:59 -0800
I learn from Robert's conference summary that the paper that won the
best of conference award is of considerable interest to me. Please will
someone give me an email contact or phone to author Christopher Thompson
of Hughes Aircraft; I would like to contact him to ask for a copy of his
manuscript. I too am essaying to construct "An Object-Oriented
Framework for Building (a) Large Distributed Ada System(s)".
Sorry I wasn't able to attend Tri-Ada this year; next year will be
better (of course!).
--
John Woodruff N I F \ ^ /
Lawrence Livermore National Lab =====---- < 0 >
510 422 4661 / v \
From: AdaWorks
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Thu, 16 Nov 1995 09:39:37 -0800 (PST)
On Thu, 16 Nov 1995, Tucker Taft wrote:
> I agree with Robert that Ada does not impose any particular
> challenge over and above the challenge of building any high-level
> language compiler for a brain-dead instruction set.
I'm not sure I would characterize the 8051 instruction set as "brain-dead."
> Certainly if you use certain Ada features like tasking, you
> will get additional out-of-line code linked in, but there were
> many Ada compilers built for 16-bit machines, and some of them
> had extremely compact run-time systems.
Yes, I am quite familiar with the 1750, a 16-bitter still in use
and widely programmed in Ada. Some of the early attempts at Ada
compilers were less than spectacuular, but current versions are
extremely good.
> As Robert points out, however, most of Ada's important features
> involve little or no additional code at run-time. Things like
> generics, private types, very strong type checking, etc., involve
> no run-time overhead.
Actually generics will involve some run-time overhead if implemented
as macro-expansions. For example, many 1750 projects do not use
generics because of the way they are implemented.
I took the time to seek out some C code for the 8051 before coming to
work this morning. Interesting to see how much of it is direct memory
accessing and register manipulation. Looking at the C code, one realizes
that the programmer was still working at a very low level of abstraction.
If we do create an Ada compiler for the 8051, one should be prepared to
understand the underlying architecture very well. I do agree, however,
that someone could design a specialized
package I8051 is ... end I8051
that would raise the level of abstraction.
> Tasking and exception handling do both
> involve out-of-line code, but not as much as you might think, and
> with protected types, there is probably little need for using
> rendezvous in a small application.
From information in Robert's earlier posting, and after thinking about this
a little more, I think there might even be room for a tasking model.
> The real point is that some processors (DSPs, 4-bit microcontrollers, etc.)
> are a pain to program in C, Ada, Modula-2, Eiffel, whatever.
> Their instructions sets are just too feeble to support any
> high-level language. However, as there is a robust market for
> 8051 C compilers, I presume the 8051 could be just as easily the
> target for an Ada compiler.
In fact, I just learned this morning that Intermetrics has/had a C
compiler for the 8051.
> I agree with Robert that it is silly
> to recommend C over Ada because somehow the instruction set implies
> it is a better fit to C. The C and Ada run-time models are extremely
> close.
This is true as long as the Ada permits the kind of low-level
programming typical of corresponding C programs. Then again, the
code I say this morning included a wierd bunch of instructions to
implement a "delay" which could be much more effective with the
Ada delay until statement. It also included some other code that
would be, as Dr. Eachus points out, prettier in Ada.
When I finally locate my old 8051 documentation, I may discover that I
have been too harsh in my criticism. Also, it seems that the current
models of the 8051 are considerably better than they were five or six
years ago -- but the fundamental architecture has not changed.
> However, one could legitimately recommend C over Ada because the C
> compiler already exists, and the Ada compiler doesn't. However, if the
> C compiler that exists is GCC, then getting a version of GNAT is simple
> enough to warrant investigation.
Even if there were a GCC port, it would also be necessary to create
several 8051 packages that specifically address the architectural
peculiarities of the beast. Now this is what one can do quite well with
Ada. Then we must understand how and when to map instructional code to
the 64KB code space and when and how to map the data to the 64KB
data space. In C, the programmer seems to do it using the same mechanisms
used by the assembler programmer.
But my sample-size of C programs for the 8051 is not very large since
most 8051 programs I have seen have been in assembler.
Richard
From: bsanden@isse.gmu.edu (Bo I. Sanden)
Subject: Re: Seeking copy of best paper
Date: Thu, 16 Nov 1995 13:19:14 -0500 (EST)
John,
We spoke at Santa Barabara conference in April. Good to "hear from you"
again! For Christopher Thompson, try cthom@iossvr.gm.hac.com or
(604) 231 3000
If it doesn't work, I'll be happy to mail you a copy of the paper.
Bo
From: Garlington KE
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Thu, 16 Nov 1995 13:00:19 -0600 (CST)
On Thu, 16 Nov 1995, AdaWorks wrote:
> On Thu, 16 Nov 1995, Tucker Taft wrote:
>
> Yes, I am quite familiar with the 1750, a 16-bitter still in use
> and widely programmed in Ada. Some of the early attempts at Ada
> compilers were less than spectacuular, but current versions are
> extremely good.
>
> > As Robert points out, however, most of Ada's important features
> > involve little or no additional code at run-time. Things like
> > generics, private types, very strong type checking, etc., involve
> > no run-time overhead.
>
> Actually generics will involve some run-time overhead if implemented
> as macro-expansions. For example, many 1750 projects do not use
> generics because of the way they are implemented.
>
Depends upon how you use the generics, I guess. The Ada flight control
software that flew in the YF-22A back in 1990-1991 used generics extensively,
using the TeleSoft TeleGen2 compiler for the MIL-STD-1750A. Particularly
when the subprograms in the generics were inlined, we got very good code as
a result. We are seeing similarly good generics code from the Tartan toolset
for the 1750.
> Even if there were a GCC port, it would also be necessary to create
> several 8051 packages that specifically address the architectural
> peculiarities of the beast. Now this is what one can do quite well with
> Ada. Then we must understand how and when to map instructional code to
> the 64KB code space and when and how to map the data to the 64KB
> data space.
I don't know how the 8051 does it, but when using an expanded memory 1750 you
have to map instructions and operands in separate logical memory spaces. This
is all done by the compiler/linker, and usually there's nothing special that
needs to be put in the code. I would assume that a GCC compiler/linker for
the 1750 would do the same things. Would it be possible that the GCC
tools for the 8051 also handle this mapping transparently to the source?
From: Garlington KE
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Thu, 16 Nov 1995 13:06:58 -0600 (CST)
On Wed, 15 Nov 1995, Robert Dewar wrote:
> Sy Wong, like Dave Weller, somehow seems to think that defining an Ada
> subset will help implementation on small machines. As long as you are
> not trying to *host* the compiler on the small machine, this thought is
> 100% bogus. Limiting the language will not in any way help the compiler
> generate code for small machines. If there are cases where limiting
> the *programs* (note the difference) helps, this can be achieved by
> use of pragma Restrictions.
>
And in fact, when he (Sy Wong) spoke at the "Why we don't use Ada" panel
session, he did mention that the set of pragma Restrictions mentioned in
the Safety and Security annex was close to what he had in mind.
Personally, I think Ada 95 would be much easier to implement on an 8-bit
machine than Ada 83. With the standard mechanisms for procedures as
interrupt handlers, protected types, access values that don't point into
a heap, etc. it should be fairly easy to write very small, very fast
programs that still can take advantage of most of the language.
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Thu, 16 Nov 95 17:24:12 -0500
" There may be such an opportunity. However, this is where I agree with
Robert rather than Sy. The Ada compiler model should be kept intact
at least for now, and we should concentrate on those market opportunities
that have a practical benefit."
This misses the point. Specifying a subset of Ada INCREASES the work to
get Ada onto small machines, and does not in anyway help, at least this
is true in the GNAT context.
Once you have a GCC port and a C compiler, then the only restriction on
the use of Ada is
(a) you can't use tasking unless you provide an implementation of the
tasking interface and have quite a bit of memory, since the
tasking runtime is not trivial in size.
(b) you can't run a program that doesn't fit. What will fit is the same
answer for C or Ada.
From: Michael Feldman
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 1995 00:01:21 -0500 (EST)
> Sy Wong, like Dave Weller, somehow seems to think that defining an Ada
> subset will help implementation on small machines. As long as you are
> not trying to *host* the compiler on the small machine, this thought is
> 100% bogus. Limiting the language will not in any way help the compiler
> generate code for small machines. If there are cases where limiting
> the *programs* (note the difference) helps, this can be achieved by
> use of pragma Restrictions.
I've followed this thread with interest and spoke at length to Sy Wong
by e-mail and at TRI-Ada.
We are not all understanding each other in these discussions. Some seem
to think a subset (cross-)_compiler_ would be desirable. I accept Robert's
statement that such a thing is undesirable and distracting.
Not everyone is understanding just what is in the compiler and what is
in the runtime. Robert's suggestion to look at pragma Restrictions is
very useful. I did that a while ago for a few of the annexes, and suggest
that we all do this if we are interested in the current problem.
One of Sy's complaints is about the _validation_ process. His understanding
- and my own - is that validation _requires_ implementation of the full
core language, and this includes the runtime, so that validation can be
done on the target.
I am puzzled now by the interaction of pragma Restrictions and validation.
Suppose, for example, that for a particular target, it becomes clear
that nested tasks - say - are simply too heavy for the
appliations to be run on that target. Therefore no program will ever
be written for that target in which that feature is used. I don;t have
the RM here, but recall a restriction in the systems programming annex
that disallows nested tasks.
The question is then whether a runtime that _assumes_ the above restrictions
can be validated. If not, is it not a bit bizarre to require a developer
to either
- forgo validation, or
- write runtime libraries supporting features that will never be used
because they are disabled by pragma Restrictions?
I would love to hear from experts like Robert and Tucker whether they
think - purely hypothetically - that it is reasonable for an implementer
to develop a runtime that assumes certain Restrictions, and essentially
tailors the compiler to enforce those Restrictions on every program.
Assuming that it is desirable for Ada implementations to be validated,
as validation provides a cedrtain predictability and "cachet", is it
reasonable to imagine a validation process that allows validation
of an implementation as described above?
I really think that the last few paragraphs sum up the 8051 issue.
I, for one, would like to understand the issues as well as possible,
and therefore will really appreciate a discussion of them that goes
beyond the "party line." Maybe I'm not the only one.
Mike Feldman
From: Michael Feldman
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 1995 01:02:40 -0500 (EST)
> pragma Restrictions has absoliutely no relevance to validation at all,
> except that if you claim to pass the relevant annexes you must pass it.
Yes, Robert, I know this. I did not ask whether it has relevance;
I asked for a discussion on whether it _should_ have relevance.
>
> On the other hand, you don't have to pass any tests which are either
> inapplicable or impractical on the target at hand. For example, Intermetrics
> passed ZERO of the I/O tests in their validation, since the Patriot nosecone
> does not have I/O.
This is not relevant to the question I asked. I specifically asked
about Restrictions, and instead of answering it, you just gave me
the party line, which I knew already.
>
> So if certain tests could not run because of capacity limitations this
> would not necessarily invalidate validation, if the AVO felt the
> exception was reasonable.
Whether or not the _tests_ run is a completely separate question from
whether practical programs will run (or should run).
>
> Obviously there are limits on this, you can't try to validate on a machine
> with 2 bytes of RAM and claim that 100% of the tests exceed the
> capacity.
I am not asking for an evasion of validation rules, I am asking for an
open discussion on whether it is reasonable to imagine a change in them.
If a feature will never be used on a given target, should its
implementation in the runtime be required for validation? This is a
different question from libraries like IO, as we all know.
>
> However, in practice most of the validation tests are small and will
> run on quite small machines without problems.
That is correct and is precisely my point. Should features be required
in the runtime if they
1. can be disabled by a Restrictions pragma (e.g., nested tasking)
2. all practical programs for the target will (should) use that Restriction
I am NOT asking whether they _are_ required; we know they are. Note:
I asked "should"?
This was obviously not an issue for Ada 83, which had no such
pragma. I am raising it as an issue with Ada 95, which does.
>
> I discussed this a while ago with Sy Wong, and I suspect he is still
> confused by pragma Restrictions, and Mike seems to have borrowed his
> confusion.
I am not confused. I'm still looking for an answer to the question I
asked. You are giving me a "party line" answer, when I asked for a
discussion about whether the party line is the best policy.
Validation is a set of rules made by humans, not Holy Writ. I realize
that the "political" answer is the one you gave me, but I was hoping for
a discussion that would go beyond politics, if that is possible.
>
> I will end the message by repeating the first sentence. pragma Restrictions
> has no relevance to validation, and in particular the issue of capacity
> limitations allowing certain tests to be ruled inapplicable is completely
> independnet of, and unrelated to, the restrictions pragma.
I will repeat my earlier sentence, namely, that is NOT what I asked.
>
Mike Feldman
From: dewar@nile.gnat.com (Robert Dewar)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 95 01:10:30 -0500
Validation is a set of rules dictated by the ISO standard. A compiler
which validated a subset in the manner you suggest would not conform
to the standard, and hence NIST would not consider the validatoin
acceptable.
I see no possible gain in persuing this idea, it was raised during the
ISO discussions of the language and soundly rejected. The reasons for
this objection seem clear enough (namely it introduces subsets
with all the confusion they bring, and has no advantage whatsoever
to implementors).
From: Bjorn Kallberg
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 1995 08:23:29 +0100 (MET)
On Thu, 16 Nov 1995, Robert Dewar wrote:
> Validation is a set of rules dictated by the ISO standard. A compiler
> which validated a subset in the manner you suggest would not conform
> to the standard, and hence NIST would not consider the validatoin
> acceptable.
Certainly not. Validation is not an ISO business. Validation is essentially
an activity defined by the US DoD, and others.
However, the ISO standard defines conforming to the standard (1.1.3). It
is clear, that a conforming implementation must implement the core language.
The question is then, if a subset (i.e. non conforming) Ada compiler
can have a commercial and user value. It seems to me that the attitude
towards subsets, additions and other experiments are not any more regarded
as totally forbidden. I think that is a good idea to be more flexible
/Bj|rn
> I see no possible gain in persuing this idea, it was raised during the
> ISO discussions of the language and soundly rejected. The reasons for
> this objection seem clear enough (namely it introduces subsets
> with all the confusion they bring, and has no advantage whatsoever
> to implementors).
From: schwarm@spectre.mitre.org (Steve Schwarm)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 95 03:12:51 EST
I find the thread of discussion very interesting because the first
Ada program I wrote was for a micro controller. It was a custom
bit slice device. The code was hand translated to "assemble" lamguage
with the ada as comments. We used ada because it allowed us to
express the data very well and the actual code was very simple. A real
compiler could have been built but there was no GNAT at that time.
Stephen(Steve) Schwarm
Principal Engineer
The MITRE Corp.
202 Burlington Rd MS G305
Bedford, MA 01730
(617)271-4600
FAX: (617)271-8140
Schwarm@mitre.org
Packet: W3EVE @ KD1CA.#RI.USA.NA
From: kst@thomsoft.com (Keith Thompson)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 1995 02:17:38 PST
> However, the ISO standard defines conforming to the standard (1.1.3). It
> is clear, that a conforming implementation must implement the core language.
In particular, AARM-1.1.3(6) says that a conforming implementation shall
> 6 Contain no variations except those explicitly permitted by this
> International Standard, or those that are impossible or
> impractical to avoid given the implementation's execution
> environment;
>
> 6.a Implementation defined: Variations from the standard that are
> impractical to avoid given the implementation's execution
> environment.
>
> 6.b Reason: The ``impossible or impractical'' wording comes from
> AI-325. It takes some judgement and common sense to interpret
> this. Restricting compilation units to less than 4 lines is
> probably unreasonable, whereas restricting them to less than 4
> billion lines is probably reasonable (at least given today's
> technology). We do not know exactly where to draw the line, so
> we have to make the rule vague.
This allows implementations *some* latitude in not supporting the full
core language; how much latitude is deliberately vague.
As always, there's nothing to prevent an implementer from providing
an unvalidated subset compiler. Such an implementer could even issue
a guarantee that the compiler passes all ACVCs except those involving
some particular unimplemented feature (say, tasking).
--
Keith Thompson (The_Other_Keith) kst@thomsoft.com (kst@alsys.com still works)
TeleSoft^H^H^H^H^H^H^H^H Alsys^H^H^H^H^H Thomson Software Products
10251 Vista Sorrento Parkway, Suite 300, San Diego, CA, USA, 92121-2718
Because I'm weird enough, I'm sick enough, and doggone it, people fear me!
From: Garlington KE
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 1995 09:56:17 -0600 (CST)
On Thu, 16 Nov 1995, Robert Dewar wrote:
> Once you have a GCC port and a C compiler, then the only restriction on
> the use of Ada is
>
> (a) you can't use tasking unless you provide an implementation of the
> tasking interface and have quite a bit of memory, since the
> tasking runtime is not trivial in size.
>From what I'm seeing here, plus the discussion of validation in the
subsequent half-dozen messages, it sounds to me that I could do the
following:
1. Get a GCC compiler for an 8051.
2. Get GNAT.
3. Port GNAT to the 8051, without tasking support.
4. Go to an AVF and say, "Dr. Dewar says in order to implement tasking,
I need quite a bit of memory on my target. I don't have it. Therefore,
the tasking tests in the ACVC are not applicable."
5. Get my certificate.
6. Provide my GNAT port to the 8051 community as a validated Ada compiler.
Is this correct?
I understand that I could also provide this compiler without validation, but
of course my potential customers will (1) ask what's wrong with it, since
they are not likely to understand the subtle distinction between "passing all
the tests" and having a validated compiler, and (2) complain that they can't
use the compiler on DoD contracts, which require a validated system.
I've only heard Sy Wong talk once, but I thought this was what he was
asking for: a compiler that implemented what the users needed, and did it
very efficiently, and cost next to nothing since the compiler developer
didn't have to spend money developing a Pthreads emulation or anything
like that.
I have a vested interest in this, of course, since once one developer
built a version of GNAT that didn't require a Pthreads-like system, that
version of GNAT could probably be ported to lots of different targets
that didn't have such a tasking interface (e.g., VAX/VMS, MacOS). I could
even get the thing validated! I couldn't use tasking, presumably, but I
could probably live with that. At least I'd have something _close_ to Ada 95.
From: Peter.Hermann@csv.ICA.Uni-Stuttgart.DE (Peter Hermann)
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 1995 17:18:56 +0100 (MET)
> >From what I'm seeing here, plus the discussion of validation in the
> subsequent half-dozen messages, it sounds to me that I could do the
> following:
>
> 1. Get a GCC compiler for an 8051.
>
> 2. Get GNAT.
>
> 3. Port GNAT to the 8051, without tasking support.
>
> 4. Go to an AVF and say, "Dr. Dewar says in order to implement tasking,
> I need quite a bit of memory on my target. I don't have it. Therefore,
> the tasking tests in the ACVC are not applicable."
>
> 5. Get my certificate.
>
> 6. Provide my GNAT port to the 8051 community as a validated Ada compiler.
... 7. and run, boy, run! ;-)
>
> Is this correct?
No.
This hurts. Are you kidding?
It was never claimed to install on a 8051 but in
a comfortable environment in order to cross to target 8051,
working there with a well optimized (both speed and size!)
executable code.
--
Peter Hermann Tel:+49-711-685-3611 Fax:3758 ph@csv.ica.uni-stuttgart.de
Pfaffenwaldring 27, 70569 Stuttgart Uni Computeranwendungen
Team Ada: "C'mon people let the world begin" (Paul McCartney)
From: Garlington KE
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Fri, 17 Nov 1995 10:36:46 -0600 (CST)
On Fri, 17 Nov 1995, Peter Hermann wrote:
> > >From what I'm seeing here, plus the discussion of validation in the
> > subsequent half-dozen messages, it sounds to me that I could do the
> > following:
> >
> > 1. Get a GCC compiler for an 8051.
> >
> > 2. Get GNAT.
> >
> > 3. Port GNAT to the 8051, without tasking support.
> >
> > 4. Go to an AVF and say, "Dr. Dewar says in order to implement tasking,
> > I need quite a bit of memory on my target. I don't have it. Therefore,
> > the tasking tests in the ACVC are not applicable."
> >
> > 5. Get my certificate.
> >
> > 6. Provide my GNAT port to the 8051 community as a validated Ada compiler.
>
> .. 7. and run, boy, run! ;-)
>
> >
> > Is this correct?
>
> No.
>
> This hurts. Are you kidding?
> It was never claimed to install on a 8051 but in
> a comfortable environment in order to cross to target 8051,
> working there with a well optimized (both speed and size!)
> executable code.
>
Actually, I'm talking about a cross-compiler for an 8051 as well. Checking
my message, I never used the word "install" in any context, and I certainly
never said the 8051 was a host.
NOW, can I do this? (If it makes it easier to visualize, let's make the host
something that runs DOS. So, I have a GCC compiler that's hosted on DOS
and targets an 8051. I have a GNAT compiler that's hosted on DOS and targets
DOS. I want to make an Ada compiler that is hosted on DOS and targets an
8051, but I don't want to have to worry about a Pthreads evaluation on an
8051.)
> --
> Peter Hermann Tel:+49-711-685-3611 Fax:3758 ph@csv.ica.uni-stuttgart.de
> Pfaffenwaldring 27, 70569 Stuttgart Uni Computeranwendungen
> Team Ada: "C'mon people let the world begin" (Paul McCartney)
From: Garlington KE
Subject: Re: Ada Waivers for Microprocessor/microcontroller applications
Date: Thu, 16 Nov 1995 13:41:06 -0600 (CST)
On Thu, 16 Nov 1995, Robert Dewar wrote:
> if it practical to have a C compiler for the 8051, then it is practical
> to have a GNAT port to it, it is that simple!
>
I thought the other requirement for a GNAT port was Pthreads support. Is
this correct, or did I misunderstand the GNAT FAQ?
From: "Robert F. Mathis" <73313.2671@compuserve.com>
Subject: Re: Mathis DRAFT TRI-Ada Trip Report
Date: 20 Nov 95 21:44:22 EST
There are a few updates I have learned about during the week after the
Conference.
The dates for the 1996 conference are only specified as "the first week of
December 1996". The sheet of paper passed out at the conference was in error.
Sunday December 1st is on Thanksgiving weekend.
While discussion on Tri-Ada 96 can occur in any forum that people like, it will
be most effective if it occurs on the SIGAda-Interest mailing list. This is the
one read by the sigada officers and the Tri-Ada 96 committee. Anyone may join
the list by sending mail to Brad Balfour (bbalfour@sw-eng.falls-church.va.us).
The mailing list is sigada-interest@sw-eng.falls-church.va.us.
-- Bob Mathis
by sztsun06.ite.iabg.de
with SMTP id <168210>; Tue, 21 Nov 1995 11:18:20 +0100
From: tonndorf@iabg.de
Subject: GNAT Validation Statement for TRI-Ada Trip Report
Date: Tue, 21 Nov 1995 11:21:17 +0100
====================================================================
Contribution to Joint Trip Report for TRI-Ada '95
On-site Validations for GNAT on Three SGI Platforms Performed
At the occasion of the TRI-Ada 95 conference the first three
GNAT implementations on the Indy, Indigo-2, and Onyx platforms were
validated for Ada 95 using the official ACVC 2.0 testsuite.
The validation was done during the conference
exhibition for GAT version 3.00 by a team from IABG, the German
Ada Validation Facility. This GNAT version will subsequently be
available for other GNAT platforms. The Ada implementations
demonstrated conforming behavior according to the Ada Validation
Procedures Vers. 4.0. However, validated status cannot be announced
before the AJPO has issued the validation certificates. This is
expected for early December.
=================================================================
Statement approved by ACT, and IABG.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+ Michael Tonndorf Tel. +49 89 6088 2477 +
+ IABG Dept ITE Fax. +49 89 6088 3418 +
+ Einsteinstr. 20 +
+ D-85521 Ottobrunn GERMANY +
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++