Re: Ada in Debian: most libraries will switch to the pure GPL in Etch

[M E Leypold <development-2006-8ecbb5cc8aREMOVETHIS@ANDTHATm-e-leypold.de>]

"Ludovic Brenta" <ludovic@ludovic-brenta.org> writes:

> I received detailed answers from AdaCore's Robert Dewar and Arnaud
> Charlet regarding the licenses of software downloaded from their
> servers.  In summary:

Ah, we'll see. Obviously the letter to me is still in transit.

> - All software downloaded from AdaCore is pure GPL, no matter what the
>   headers say.

> - This also applies to software downloaded from the CVS server in
>   source-only form.

So, if I get the source from ACT CVS it's GPL and if I get it from one
of the Archives somehwere out there, it has GMGPL. I'm talking GtkAda
2.2.1 and GtkAda 1.x specifically where there is strong indication
that it has been distributed as GMGPL (by whomever ...).

> - They refuse to give any assurances regarding copyright ownership, so
>   I feel that I now need to go ask the authors.

Please. Also It would be useful to ask AdaCore to document the
provenience of the sources they distribute: At which distribution
points the got the source, what are their assurances that they are
allowed to distribute the sources under GPL and so on :-). What their
baseline version is (that is from which version on they started to
"contribute" to the code base).

> - They will not sign a license document, even if a lawyer asked them
>   (they are not required to sign anything, of course).

I think they are required to say very clearly on which basis they
think they have the right to distribute the software they haven't
completely written themselves (which applies to most from libre2
AFAIS).

> As you all know, Gentoo has switched to the pure GPL for all Ada
> libraries, in accordance with these statements.  

> IANAL but this seems to be the only legal route, so Debian will
> follow suit and switch to the pure GPL for future versions of the
> libraries.


> ASIS-for-GNAT is copyright (c) Free Software Foundation.
> Theoretically, I could ask the FSF for a GMGPL license, and maybe get
> the sources from gnat-asis.sourceforge.net, but that's unlikely to
> work with GCC 4.1 without major work.  So, switch to pure GPL.

As far ASIS is concerned: I consider that to belong to the tool
sector, so GPL is no problem. ASIS is, in a sense an extension of
GNAT. I did not grok all of you statement here, though: What exactly
is the situation wrt to ASIS: 

> AUnit was written by Ed Falis, an AdaCore employee.  Only source is
> AdaCore, so switch to pure GPL.

Also a tool, wouldn't get linked into an executable for the customer:
In my eyes GPL is OK here.

> AWS was written by Dmitryi Anisimkov and Pascal Obry.  Neither of them
> being an AdaCore employee (at least AFAIK), there may be a way to
> acquire a GMGPL license from them.  They may also decide to fork the
> project on a new repository.  Pascal's site [1] has a download page
> that points to AdaCore's web site.  In the mean time, AdaCore is the
> only source, so switch to pure GPL.

Older Versions where GMGPL AFAIK. We should ask the authors for either
a GMGPL version (from their archives) or at least for the last GMGPL
version.

> Florist was written by Florida State University; it is possible to get
> a version ported to GCC 4.1 from http://gnat-florist.sourceforge.net,
> so I'll do that.  GMGPL.

This is, I think the same version as 3.15p as also available from
Baker's site. Since ACT did hardly anything with their GPLed florist
apart from removing the linking exception and adding a copyright
statement of their own, I'd judge that version still good (after all:
POSIC hasn't changed, it's mostly a question how it will build on
various platforms).

> GLADE was written jointly by AdaCore, the ENST, A. Strohmeier, T. Wolf
> and J. Kienzle.  Only source is AdaCore, so switch to pure GPL.

I agree as far as the tools themselves are concerned. But I can get a
version of glade-3.15p from non-ACT servers and (if we really want to
start splitting hair then: here) I can see linking exceptions in a
number of files. Perhaps these are exactly the runtime that gets
linked to client and server. This aspect is exactly why I'd really
want to know more about the copyright of single files in that whole
mess.


> GtkAda was written by Emmanuel Briot, Joel Brobecker, Arnaud Charlet,
> and Nicolas Setton, who are all AdaCore employees.  Only source is
> AdaCore, so switch to pure GPL.

So there has never been a GMGPL version? Never ever? I find that
odd. I suppose you can now never get Emmanuel Briot, Joel
Brobecker, Arnaud Charlet to admit that there was ever a GMGPL
version?

As far as past versions are concerned, I pull the following Texts from
my archive, which have accomanied GtkAda on libre (the old one) for
some time:

  Version 1.3: This package is distributed under the GPL license,
  slightly modified so that you can create proprietary software with
  this toolkit. The license is actually the same as the GNAT library
  itself. You should also read the Gtk license itself if you intend to
  do proprietary software based on gtk and GtkAda.

  Version 2.0.0: This package is distributed under the GPL license,
  slightly modified so that you can create proprietary software with
  this toolkit. The license is actually the same as the GNAT library
  itself. You should also read the Gtk license itself if you intend to
  do proprietary software based on gtk and GtkAda.

  Version 2.2.1: This package is distributed under the GPL license,
  slightly modified so that you can create proprietary software with
  this toolkit. The license is actually the same as the GNAT library
  itself. You should also read the Gtk license itself if you intend to
  do proprietary software based on gtk and GtkAda.

  Version 2.4.0: This package is distributed under the GPL license,
  slightly modified so that you can create proprietary software with
  this toolkit. The license is actually the same as the GNAT library
  itself. You should also read the Gtk license itself if you intend to
  do proprietary software based on gtk and GtkAda.


I suppose if I just get the version of GtkAda from my archive, I'll be
covered by GMGPL, or what? Or if I get this one:
http://www.adapower.net/libre/gtkada/GtkAda-2.0.0.tgz

ACT didn't even bother to change the source of 2.4.0 distributed from
libre2. The just somehow lost (not even substituted) the license
notice when they started to mirror libre.act-europe.fr.

All this has a positively Orwellian touch (especially if I factor in
the now well known Robert Dewar quote): Changing the past
retroactively.



> libgnat is GMGPL, since we obtained it from the FSF's repository as
> part of GCC 4.1.

Yep.

> PolyORB is copyright (c) Free Software Foundation.  The situation is
> exactly the same as with ASIS: the FSF doesn't seem to be aware of
> PolyORB, the only place to get it from is AdaCore, and that's pure
> GPL.  PolyORB has a home page [2], but the download link points to
> AdaCore.  (that doesn't matter for now, since Debian does not provide
> PolyORB).

polyorb/README

   -- As a special exception,  if other files  instantiate  generics from this --
   -- unit, or you link  this unit with other files  to produce an executable, --
   -- this  unit  does not  by itself cause  the resulting  executable  to  b  --

This even in the version obtained from ACT.

Other version can be gotten from FTP-Servers /= ACT's. 

If I go to http://polyorb.objectweb.org I find "As a special
exception, if other files instantiate generics from this ..." and 

"PolyORB 1.3r (2005-06-15) (...) This is the latest release of
PolyORB.  Release is a snapshot of current research work on middleware
architecture carried out by the ENST and the LIP6-SRC. (...) Note:
this snapshot is distributed on Libre's page."

It seems the copyright somehow gets lost in between objectweb.org and
ACT. I find it hardly credible that the agreement (of whatever kind)
between objectweb and ACT is, that ACT distributes under "pure
GPL". I'd suggest that the community ask objectweb for either
subversion access or the last GMGPL snapshot.


> Templates_Parser, a library that is incorporated into both AWS and
> GPS, is copyright (c) AdaCore.  It seems that the author, Pascal Obry,
> has assigned copyright to them.  If that's untrue (and the headers are
> in error), 

ACT is in the habit of rewriting headers.

> then my remarks on AWS apply to Templates_Parser, too.  (Debian
> includes Templates_Parser as part of AWS and GPS, but not as a
> separate package).


> XML/Ada was written by Emmanuel Briot (an AdaCore employee), ...
> Christophe Baillon and Martin Krischik.  Only source is AdaCore, so
> switch to pure GPL.

From the xmlada README: 

  "As a special exception, if other files instantiate generics from ..."

ACT might be in their rights to strip the linking exception from all
source they distribute from their site. But since at least a part of
that software got distributed under different licenses earlier, I
doubt they can take back the linking eception on all that copies
floating around on various backup media and ftp servers. Not
retroactively.

That also applies to all the other packages where the copyright is now
being stripped retroactivly.

> AdaCore, or AdaCore employees, are authors of AUnit, GLADE, GtkAda,
> and XML/Ada.  They might volunteer a GMGPL version of their work, but
> I'm not going to ask them for one.

I understand that. Is BTW, anybody here who's intrested in
constituting a GMGPL baseline of some core libraries?


> Since GtkAda has been mentioned more than any other library, I'd like
> to point out that Debian already includes two alternatives:

Ludovico. I understand that. But: Gtk+ has been LGPL for very good
reasons and part of the success of Gnome is based on that. In my
opinion the future on Unix/Linux is Gtk+/Pango/Cairo. And it comes as
a special bonus that these libs are portable. Everything else is
probably good to hack some tool now, but not to build something that
can survive in for the next 10 years. 

I'm not talking about a single program but about a code base. As I
very small example, I've developed (among other things) a flowing
label widget (Text in label flows on resizing) for GtkAda. I wouldn't
see any sense in developing (and releasing to the community)
infrastructure of that kind for what I consider a niche solution.

> - AdaBindX [3], a binding to X11 and LessTif by Hans-Frieder Vogt
>   which is under GMGPL.  X11 is under X11 (MIT) license, and
>   LessTif[4] is under LGPL.  But AdaBindX has not been updated since
>   2000, and is not portable.  


> And not as good-looking as GTK+.

>   According to Debian's Popularity Contest [5], this package has zero
>   users, so I am tempted to drop it from Etch.  Speak up if you want
>   me to keep it.

Just announce it again when you drop it. A X11-Binding would have been
nice for experiments, especially for trying to write a Ada native
widget set, but that not here, rather science fiction (or more likely
alternate history ...), so ...

> 
> - TASH, the Tcl Ada SHell [6], includes a binding to Tk; it is
>   available as a Debian package, under GMGPL, from Ada-France [7] and
>   AdaWorld [8].  TASH is portable, but has not been updated since
>   2003.

> Thoughts, comments, offers to help?

I'd like to help establishing GMGPL libraries (for libraries that get
linked into distributed executables, GPL is OK with me for the tool
chain). 

My thoughts, rather jumbled, I fear I've offered above.

Sorry for all this. I feel like single handedly killing Ada on Debian
for asking that stupid licensing question at the beginning.

Regards -- Markus