Ada in Debian: dropping support for the Distributed Systems Annex

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The package gnat-glade in Debian has been providing support for Annex E
(the Distributed Systems Annex) since 1999. With this package, it is
possible to write programs that are distributed over many computers.

I have found a critical bug that makes the version currently in testing
(gnat-glade 2007) unusable for real-world applications. The problem is
described in detail in [1]. Now that AdaCore has abandoned GLADE in
favour of PolyORB, there will be no solution to this problem.

Therefore, starting in April 2008, I packaged PolyORB 2.3 for
Debian. During testing, I discovered other problems [2] that make
PolyORB unusable too. Consequently I decided not to upload the package
to Debian, but to wait for the next release.

After AdaCore released PolyORB 2.4 (as part of GNAT GPL 2008 Edition)
I resumed work but the problems are still there, whether using gnat-4.3
or GNAT GPL 2008 as the compiler [3].

As a consequence, I am sorry to announce that Debian 5.0 "Lenny",
scheduled for release in September 2008, will not support the
Distributed Systems Annex.

I will continue working on PolyORB in Debian as my time allows. If you
think you can help, please do not hesitate to get in touch with me. The
packaging scripts are on Ada-France's public monotone server [4,5].
Maybe one day I will be able to integrate a rock-solid PolyORB in some
future version of Debian.

[1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=492100
[2] see thread started by :
http://lists.adacore.com/pipermail/p...ay/000950.html
[3]
http://lists.adacore.com/pipermail/p...ne/000984.html
[4] http://www.ada-france.org:8081/branc...debian.polyorb
[5] http://www.ada-france.org/article131.html

- -- Xavier Grave (Thanks to Ludovic Brenta for his help on all of this)


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkiNaXsACgkQVIZi0A5BZF4rOACfchNTQaosh0 kGZ5nQcA5m0g5G
imMAn3ZxIKdm77bobTGatLJbFES507FI
=3gdX
-----END PGP SIGNATURE-----

On Mon, 28 Jul 2008 08:38:56 +0200, xavier grave wrote:

....
> As a consequence, I am sorry to announce that Debian 5.0 "Lenny",
> scheduled for release in September 2008, will not support the
> Distributed Systems Annex.
>
> I will continue working on PolyORB in Debian as my time allows. If you
> think you can help, please do not hesitate to get in touch with me. The
> packaging scripts are on Ada-France's public monotone server [4,5].
> Maybe one day I will be able to integrate a rock-solid PolyORB in some
> future version of Debian.

Thank you very much Xavier, Ludovic and others for your efforts on this!

I use Annex E/GLADE heavily in my (unpublished) financial analysis/stock
trading software. The early releases of the system worked OK once
you knew of its limitations. I'm using GNAT GPL 2006 with GLADE, and
that's OK too (but has "issues"). It's been fun and the tools have
shown great promise. So I am disappointed that it's not become more
popular and better supported.

In these days of distributed systems and multicore processors, running
complex applications, you'd think Annex E support would be universal.
What's happening here? It doesn't seem to have been used in any
major application (Air Traffic Control, Baggage Handling etc).
Obviously ACT isn't recommending it, but why isn't there customer pull?

From my own perspective, it's not so critical, since what I have works,
and I am moving on to new pastures now anyway. I had a look at
PolyORB, but that seemed incomplete for DSA when I tried it.
I'll test PolyORB again if I hear of progress with it, if time allows,
but the "market" is telling us something :
"we don't need or want Annex E"

Thanks again guys for your heroic efforts.
--
Dr. Adrian Wrigley, Cambridge, UK.

On Mon, 28 Jul 2008 17:49:21 GMT, Dr. Adrian Wrigley wrote:

> In these days of distributed systems and multicore processors, running
> complex applications, you'd think Annex E support would be universal.

Well, maybe RPC is too heavy-weight, and problematic for heterogenous
systems. It represents a client-server, synchronous (OK, there is APC too),
peer-to-peer, static architecture. Procedural abstraction (RPC) is both
low-level and expensive. Annex E lacks remote objects, remote types,
agents, routing, time synchronization, abstracted networking hardware
support.

> What's happening here? It doesn't seem to have been used in any
> major application (Air Traffic Control, Baggage Handling etc).
> Obviously ACT isn't recommending it, but why isn't there customer pull?

For the distributed systems we are designing, we just deploy other
architectures.

And AdaCore seems to look to the other side too:

http://www.adacore.com/2007/12/08/wo...ng-middleware/

--
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Yes.
I would rephrase it to state SAAB has requested RTI to provide
Ada-bindings for NDDS and that is a fairly thin binding (Quite close to
OMGs IDL2Ada specefication)

However the cost for NDDS is quite high at the moment we are talking
about figures in the magnintude of $1000 + runtime per target CPU.

However DDS is a very powerfull tool for data distrubution.

/Per

Dmitry A. Kazakov wrote:
> On Mon, 28 Jul 2008 17:49:21 GMT, Dr. Adrian Wrigley wrote:
>
>> In these days of distributed systems and multicore processors, running
>> complex applications, you'd think Annex E support would be universal.
>
> Well, maybe RPC is too heavy-weight, and problematic for heterogenous
> systems. It represents a client-server, synchronous (OK, there is APC too),
> peer-to-peer, static architecture. Procedural abstraction (RPC) is both
> low-level and expensive. Annex E lacks remote objects, remote types,
> agents, routing, time synchronization, abstracted networking hardware
> support.
>
>> What's happening here? It doesn't seem to have been used in any
>> major application (Air Traffic Control, Baggage Handling etc).
>> Obviously ACT isn't recommending it, but why isn't there customer pull?
>
> For the distributed systems we are designing, we just deploy other
> architectures.
>
> And AdaCore seems to look to the other side too:
>
> http://www.adacore.com/2007/12/08/wo...ng-middleware/
>

On Jul 28, 7:49 pm, "Dr. Adrian Wrigley"
<a...@linuxchip.demon.co.uk.uk.uk> wrote:
> On Mon, 28 Jul 2008 08:38:56 +0200, xavier grave wrote:
>
> ...
>
> > As a consequence, I am sorry to announce that Debian 5.0 "Lenny",
> > scheduled for release in September 2008, will not support the
> > Distributed Systems Annex.
>
> > I will continue working on PolyORB in Debian as my time allows. If you
> > think you can help, please do not hesitate to get in touch with me. The
> > packaging scripts are on Ada-France's public monotone server [4,5].
> > Maybe one day I will be able to integrate a rock-solid PolyORB in some
> > future version of Debian.
>
> Thank you very much Xavier, Ludovic and others for your efforts on this!
>
> I use Annex E/GLADE heavily in my (unpublished) financial analysis/stock
> trading software. The early releases of the system worked OK once
> you knew of its limitations. I'm using GNAT GPL 2006 with GLADE, and
> that's OK too (but has "issues"). It's been fun and the tools have
> shown great promise. So I am disappointed that it's not become more
> popular and better supported.
>
> In these days of distributed systems and multicore processors, running
> complex applications, you'd think Annex E support would be universal.
> What's happening here? It doesn't seem to have been used in any
> major application (Air Traffic Control, Baggage Handling etc).

At Eurocontrol CFMU (where I work), the software has been distributed
since before Annex E (or Ada 95) even existed. Moreover, some systems
are written in languages other than Ada. These are two reasons for not
using Annex E in that context. I cannot speak for others, of course.

> Obviously ACT isn't recommending it, but why isn't there customer pull?

I wouldn't go so far as to say AdaCore isn't recommending Annex E.
They've simply switched their implementation from GLADE to PolyORB.
So, AdaCore "obviously" recommend (and commercially support) GNAT Pro
+ PolyORB. I'd be interested in feedback from users of PolyORB for
Annex E, in particular what version of GNAT or GCC they use, since the
problems Xavier found were impedance mismatches between PolyORB, the
compiler and the run-time library.

[...]
> "we don't need or want Annex E"

That's what I gather from the fact that Xavier is the only one to my
knowledge interested enough to spend hours upon hours of his time
trying to get Annex E to work. I think that with enough interest (i.e.
man-hours), GCC 4.3 could have been made to work with either GLADE or
PolyORB.

> Thanks again guys for your heroic efforts.

I don't deserve these thanks, so I'll pass them on to Xavier

--
Ludovic Brenta.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

> Therefore, starting in April 2008, I packaged PolyORB 2.3 for
> Debian. During testing, I discovered other problems [2] that make
> PolyORB unusable too. Consequently I decided not to upload the package
> to Debian, but to wait for the next release.
>
> After AdaCore released PolyORB 2.4 (as part of GNAT GPL 2008 Edition)
> I resumed work but the problems are still there, whether using gnat-4.3
> or GNAT GPL 2008 as the compiler [3].
>
> As a consequence, I am sorry to announce that Debian 5.0 "Lenny",
> scheduled for release in September 2008, will not support the
> Distributed Systems Annex.

It means, It doesn't even work it to give a try compile it under debian?

Sébastien
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Cygwin)

iD8DBQFIkHkV+zV9xm4PlDQRAjkRAJ4/hjbswBL3+6yPg/y6zznkL1K6+gCeItuQ
B70JkuVAimrC9y6fXcm+oJA=
=1xcX
-----END PGP SIGNATURE-----

Sébastien Morand wrote:
> Xavier Grave wrote:
> > As a consequence, I am sorry to announce that Debian 5.0 "Lenny",
> > scheduled for release in September 2008, will not support the
> > Distributed Systems Annex.
>
> It means, It doesn't even work it to give a try compile it under debian?

gnat-glade itself compiles (there is a binary package in Debian
testing) but if you try to compile your own distributed programs with
it, you get compiler errors due to version mismatches between compiler
(gnat-4.3), run-time library (libgna-4.3) and libgarlic-dev. These
mismatches are detected by means of the .ali files.

The state of PolyORB is that both PolyORB itself and the distributed
program compile fine, but then the distributed program gives errors at
run time.

--
Ludovic Brenta.

After modifying GLADE-2007, to compile in GNAT-2008, I now, have only
a single problem in compiling one example program, but I also, had the
same error when I compiled the same example in GLADE-2007 with
GNAT-2007.

And all partitions that I have adopted to use either 2007 or 2008 GNAT
system works as expected.

As for Polyorb, the examples work fine. But may not do much more testing,
because Polyorb is too much a JAVA based system for me.





In <3b572050-4e27-4316-89d3-945333c3ec8e@l42g2000hsc.googlegroups.com>, Ludovic Brenta <ludovic@ludovic-brenta.org> writes:
>S=E9bastien Morand wrote:
>> Xavier Grave wrote:
>> > As a consequence, I am sorry to announce that Debian 5.0 "Lenny",
>> > scheduled for release in September 2008, will not support the
>> > Distributed Systems Annex.
>>
>> It means, It doesn't even work it to give a try compile it under debian?
>
>gnat-glade itself compiles (there is a binary package in Debian
>testing) but if you try to compile your own distributed programs with
>it, you get compiler errors due to version mismatches between compiler
>(gnat-4.3), run-time library (libgna-4.3) and libgarlic-dev. These
>mismatches are detected by means of the .ali files.
>
>The state of PolyORB is that both PolyORB itself and the distributed
>program compile fine, but then the distributed program gives errors at
>run time.
>
>--
>Ludovic Brenta.

Point of information: Glade has reappeared in GnatPro 6.1.2

I guess this means that there /are/ paying customers who use Annex E...

--
---------------------------------------------------------
J-P. Rosen (rosen@adalog.fr)
Visit Adalog's web site at http://www.adalog.fr

anon a écrit :
> As for Polyorb, the examples work fine. But may not do much more testing,
> because Polyorb is too much a JAVA based system for me.

PolyORB... Java based... Lot of nonsense there!

It really looks like "anon" is a shared account. Sometime "anon" post
some interesting messages and sometime it is just nonsense... Sorry
"anon" but that's really a bad netiquette! You should either post with
your real name or stop posting stupid messages on this news group.

Pascal.

--

--|------------------------------------------------------
--| Pascal Obry Team-Ada Member
--| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
--|------------------------------------------------------
--| http://www.obry.net
--| "The best way to travel is by means of imagination"
--|
--| gpg --keyserver wwwkeys.pgp.net --recv-key C1082595