Re: Open Systems closed to Ada?

[cis.ohio-state.edu!zaphod.mps.ohio-state.edu!usc!enterpoop.mit.edu!linus! linus.mitre.org!news!emery@ucbvax.Berkeley.EDU (David Emery)]

<well, I'm breaking my promise again to not post on this, but here goes.>

The goals of the P1003.5 effort was to provide Ada with access to the
facilities provided by a POSIX-compliant operating system.  A similar
group, P1003.9, worked on a FORTRAN binding to POSIX.  It's worth
noting that Unix has been multi-lingual for years, starting with
FORTRAN, and continuing through lots of other languages.  There is
substantial experience in the Unix community with providing access to
Unix services from languages other than C.  

>>When I got started in the P1003.5 effort, I was hoping that we would
>>be welcomed by the rest of POSIX.  What I expected was disinterest.
>>What I found was hostility.  

>Why is this surprising?  Here are these folks who have been cookin' along
>with UNIX for, say, eight years or so.  They know they need a standard;
>they're getting desperate.  It is academically, intellectually, and most
>important, commercially imperative that they get a standard they can use.
>The definition of the system interface is cast, as it has been for many
>years, in terms of C.  They're ready to do it; they know what they need;
>they've finally hashed out most of their differences...and all of a sudden,
>along comes the kid brother saying "hey, wait for me!  You gotta let me
>play too!  Mom said so!"

>Why should they be patient with someone joining the game late, trying to
>get a share of the action?  Why *shouldn't* they be hostile?  They've got
>work to do too.  The attempt to shift to language-independent definitions
>of interfaces late in the game was a major obstacle to the work.

I strongly suspect that you have not been a participant in the POSIX
standardization efforts.  Ada (and FORTRAN) binding efforts have had
little or no negative effect on the development of the C
standards/bindings.  During the balloting of POSIX P1003.1, a couple
of us sent in ballots commenting on some specific areas from an Ada
perspective.  In some cases, these ballots were used to decide between
two otherwise equal approaches.  During the development of the P1003.4
and P1003.4a real-time and threads standards, the Ada balloters have
been much more involved, but in this case there is not the base of
experience that existed for P1003.1, and the Ada community has
substantial experience implementing concurrency both on top of
traditional Unix systems, and also for real-time applications.

What the P1003.5 working group wanted from the POSIX standardization
effort was a reasonable forum to work in.  In particular, we wanted to
make sure that the Ada binding was technically correct from the POSIX
side, and also that it was usable from the Ada side.  During the
development and balloting of the standard, we received some really
outstanding ballots from balloters with little or no knowledge of Ada,
but substantial experience with C and POSIX.  It's probably worth
noting that P1003.5 had the _largest_ balloting group of any POSIX
standard to date, and achieved 91% consensus within that group, well
above the average approval rating for a POSIX standard.  

My problem comes with the conduct of the standardization committees.
As has been noted previously, ISO WG15, and also the IEEE TCOS SEC,
have been adept at developing bureaucratic roadblocks for language
bindings other than C.  Ada is not alone in this respect; the FORTRAN
group (P1003.9) has been similarly obstructed.  Much of this is
centered around the call for "language independence".  ISO has ruled
(for better or worse, mostly worse, in my opinion) that POSIX
standards should be developed using a language-independent notation,
with "thin" language bindings.  This has been used as justification to
prevent the FORTRAN and Ada bindings from achieving standardization,
but the same line of reasoning has NOT been applied to equivalent C
bindings.  

The net effect of these procedures has been to permit work on C
bindings to go forward, while holding back other language bindings
from full standarization.  I find this objectionable.  If a
requirement like language-independence is established, it should be
applied equally.  If language-independence and thin bindings are a
good thing for Ada and FORTRAN, then they are a good thing for C, too.
If they're not a good idea for C, why does anyone think they are good
for other languages?  

I don't want to see _any_ language binding prevented from achieving
standarization.  But, as an Ada advocate and Ada bindings developer, I
have a personal stake in the standardization of Ada bindings to open
systems standards such as POSIX, PHIGS, and X Windows.  

				dave