On Fri, 5 Sep 1997, Michael Feldman wrote:

> [Peter said]
> >
> > > Hearing from Robert Dewar, that DOS will be ignored
> > > in future versions of GNAT beyond 3.07
> > > I am concerned about the spreading of Ada:
> > >
> > > At least GNAT3.10 should be ported to DOS in order to
> > > have a relatively high reliable compiler there.
> > >
> > > grass roots are important.
>
> [David answered]
> >
> > I'm not sure Robert reads this mailing list, but I think he'd merely
> > point out that GNAT 3.10 could definitely be supported in DOS,
> > _provided_ 1) Somebody paid ACT to do it, or 2) Somebody used the GNAT
> > 3.10 public sources/DJGPP to build a version for DOS.
> >
> [Mike jumps in]
>
> Peter, have you communicated personally with the ACT folks on this?
> I have. If it were easy to build 3.10 for DOS (or even 3.09), it would
> have been done long ago. Building the compiler is easy; building the
> runtime is not, especially the tasking support. What makes it worse is
> the dependence of the GNAT/DOS runtime on DJGPP, and subtle interactions
> (or even DJGPP bugs) are really at the root of the problem.
>
> DJGPP is also a free-software undertaking, relying a lot on volunteer
> effort. To resolve the lingering unpleasant surprises in GNAT/DOS
> will apparently require a lot of effort from both groups.
>
> The typical first-year student comes in with, or buys, a nice PC
> these days, either a Mac or a decent Pentium. MachTen CodeBuilder
> is an _excellent_ $99. Unix, with an up-to-date, supported GNAT.
> The compiler is free of course; the $99. is for the Unix.
>
> For all that I am a Mac fan, I recognize that on a decent hardware
> platform, Win 95 is an acceptable (if not wonderful) piece of work.
> Moreover, Linux is readily available with a perfectly good, up-to-date,
> GNAT. So for the student community whom I consider my "constituency",
> DOS is not much of an issue anymore.
>
> I've observed that the Ada community is bursting with good ideas,
> but many of the idea-generators don't have the time or inclination
> to implement their own ideas. Yes, it would be nice to continue
> GNAT for DOS. The sources are public (at least the 3.09 sources,
> pending release of 3.10). I am aware of some of the problems
> encountered in building the 3.09 runtime, which is why no 3.09 was
> released. Perhaps someone - or some small group - out there wants
> to try again, and try to work with the DJGPP folks to iron out the
> interactions. Peter, how 'bout you?
>
> Mike Feldman

[and John continues to weave...]

While a 3.09 or 3.10 release of GNAT for DOS would be nice, what
is so "wrong" with the 3.07 release of GNAT for DOS?  I seriously
doubt people or companies are using GNAT for DOS as a hardcore
software production platform.  (Hence the lack of paid ACT support
to further GNAT for this OS... You get what you pay for!) There
may still be a lot of "DOS" machines out there - but you have to
admit that DOS is beginning to show its age. ("I'm not dead
yet!"... ;-)  GNAT 3.07 for DOS works fine - especially for someone
just starting out with Ada...  I find its quality more than
acceptable.  Are you suggesting that 3.07 is not a "relatively
high reliable compiler"?  I seriously doubt that the delta in
"quality" between 3.07 and 3.10 is such as to dissuade someone
from trying Ada out!

As Mike mentioned, another real show stopper here is that DOS really
wasn't designed (from an OS standpoint) to support things like tasking
(Ada's RTS) - hence the DJGPP add-on and all the headaches it
introduces to a GNAT DOS release....

As far as the "grass roots" movement idea goes, I'd say that Linux
has pretty much been running with that baton for the last
couple of years...  (I actually belong to the OS/2 camp though...)

My $0.02.

--John Norcross