Re: Is the "Ada mandate" being reconsidered?

[jsa@organon.com (Jon S Anthony)]

In article <4pn0rs$mbe@gde.GDEsystems.COM> Tom Robinson <robinson@gdesystems.com> writes:

> >Well, this sounds good, it's just wrong.  Gnat is free and of better
> >overall quality than all but the very best Ada83 compilers.
>
> I am interested in what measure you are using for quality.  From the
> small snipits I have read on the net it is not clear that the
> generated code quality of gnat is quite up to current standards yet.
> Am I missing something here?  Has anyone done ACEC or PIWG
> comparisons yet?

I am speaking of correctness, robustness, flexibility, fullness of
implementation, error messages and error recovery, integration with
widely available and common tools, and speed of compilation.  Actually
in terms of speed I don't think any of the Ada83 compilers can even
come close (well, discounting certain incremental compilation
situations in something like Apex).  No way.

The generated code quality is often about the same as for GCC C
compiler.  This is really quite good, but there are Ada constructs at
present which can derail things.  But this was true of Ada83 compilers
as well.  For me that is "good enough" for now.  If you are in some
sort of "hard real-time" situation where the code has to be as good
as possible, then I would be very careful or looking elsewhere.


> >Thomson's ObjectAda compiler is dirt cheap for personal
> >use and the professional version is cheaper than a "professional" C++
> >package.  Also, Gnat is on all sorts of platforms.
> 
> Is it really?  When I look at the Ada 95 validated compiler list it 
> looks pretty small to me.  So you're saying that gnat is available as
> long as I am willing to pay for a validation and arrange for maintenence
> or do it myself.  

Yes, really.  Who cares about validation?  Maybe you do, but I don't.
I _like_ the fact that there is an ACVC for all sorts of QA reasons.
But the actual stamp means nothing to me.  I'll tell you a little
secret: C/C++ programmers don't care about AJPO validation
certificates either.  For me (and I am sure most "usual" commercial
developers), if I knew the thing passed the test suites (ACVC and any
internal ones - especially ones that _I_ submitted) on _any_ common
platform, that would be quite sufficient.  And even if it failed a
few "goofy" ACVC tests, that wouldn't break my heart either (even
though that would prevent an official stamp).




> >And since
> >ObjectAda uses the Intermetrics AdaMagic frontend, it will likely be
> >all over the place too (or at least highly compatible counterparts
> >based on the same frontend).
> >
> >No, overall, this situation is _vastly_ better today than before.
> >
> >
> 
> I don't understand how you can say that the situation is _vastly_ better
> based on what you have stated.  It is a fact that the Ada 83 compiler
> choices have been shrinking.

I can state it very easily because it is true.  People can use obtain
and use very high quality compilers for all sorts of common platforms
(PCs in particular) for little or nothing.  In my book that is indeed
_VASTLY_ better than 5 or 6 years ago where you had it on only a
couple common platforms and to get it would cost a fortune.  I certainly
would not be using it if the situation were as before - I just plain
couldn't!!


>  All one has to do is look at the list of Validated compilers and
> realize that many companies that did Ada 83 validations (Alsys,
> TeleSoft, Verdix, Meridian, Systeam, (off the top of my head)) are
> no longer operating under those names.  They have merged with other
> companies or gone out of business

So?  Who cares?  The same thing is happening in the compiler biz no
matter the language.  The question is, are the products of today
better fitting into the market place?  From where I sit the answer
isn't just "yes" it's "HELL YES!"


> announced that they will use non-proprietary front ends for their
> product offerings (DEC).

One hell of a good idea, if you ask me.  A smart idea which allows for
putting true value added where it means something - not just
duplicating the same effort over and over again.  Intelligent reuse
with continual improvement!  Imagine that - in software!  Go figure!
BTW, where do you think most of the C and C++ compilers have come
from?  Hmmmm?


>  Other OEMs have transitioned from directly
> offering Ada products to offering them through a 3rd party (IBM->OCSystems).

Great idea.  Excellent idea.  Why the hell should IBM (a _hardware_
company) be making Ada (or any) compilers?

 
> I think that the Ada market is in transition.  As an Ada user the
> transition to Ada 95 seems like it might be a bit tricky.  For large
> projects it looks like there are going to be more decisions that I
> (as a buyer of Ada technology) will need to make:

Sure, it is in transition.  And that is a VERY good thing.  If it
weren't it would be dead.  I don't really understand your concerns
here, but I'm sure that for you they are legitimate.

For us, the decision was obvious.  One of the things we did here was
move a "large" (around 200KSLOC) legacy '83 program over to Ada95 with
GNAT (and yes, we have ACT support - which BTW has been much better
than what I am used to from other software vendors - be they compiler
shops or whatever).  This was from a Verdix based project.  The
situation is now so much better than before, that it is simply
wog-boggling.  We can and we have moved it to several other platforms
- including Win95.

No, for the Ada world, the situation is so much better than a few
years ago, that it is positively stupefying.


/Jon
-- 
Jon Anthony
Organon Motives, Inc.
1 Williston Road, Suite 4
Belmont, MA 02178

617.484.3383
jsa@organon.com