Re: what DOES the GPL really say?

[dewar@merv.cs.nyu.edu (Robert Dewar)]

<<I'm curious. What say had the DoD in GNAT? Was the possibility of
experimental activity with the compiler the only reason the DoD insisted
on having the sources distributed under the GPL?

There is a certain economic sense to goverment support of free software as
has been argued in gnu.misc.discuss recently; in this connection, it would
be useful to know what the DoD's take on the GPL was.>>


The DoD funded GNAT to provide a research vehicle for academic use, and it
was clear for this purpose that free availability of sources was important,
so the NYU contract with the DoD insisted on

   a. The use of the GPL for all sources
   b. The assignment of the copyright to the FSF

The second part is important, because it ensures that GNAT remains under
the GPL. Note that the copyright holder of GPL'ed software can always
change their mind, and not use the GPL on a subsequent release. This
for instance could have happened if NYU had held the copyright, and had 
decided to try to commercialize GNAT after the contract ended. But the
assignment to the FSF involves a commitment for continued availability
under the GPL.

The narrow focus of the DoD interest in GNAT partly resulted from dealing
with rather fierce opposition to the entire project from some, but by no
means all, vendors of proprietary Ada products. Indeed the contract with
the DoD was for a subset of Ada 95, excluding some key feaures (including
fixed-point and subunits, that were perceived to be required in the market
place, but not critical for academic research -- validation was also
excluded).

In fact the missing features were included anyway, but not on government
time. Both Ed and I took sabbaticals during the project, so we could work
on it on our own time, not being paid by government sources, so we used
(part of) this time to implement the missing features.

We always recognized the potential of GNAT in two other areas

    1. Teaching. The DoD did not have the view of GNAT as useful for
       this purpose, and indeed Mike Feldman's proposal to use GNAT
       for the academic Ada compiler to be supported by the DoD, was
       turned down, despite its advantageous price, on the grounds that
       ACT was not a credible organization for carrying out the validation
       (a bit ironic in retrospect, given that ACT is still the only
       company to have done 100% validations with all the annexes, but
       it's easy to be wiser after the fact -- of course it is a bit of
       a puzzle why validation was required at all in this project).

       Nevertheless, we have always regarded this as a major opportunity
       for GNAT, and have put a lot of work, e.g. particularly in the
       generation of good error messages) to facilitate this use. We
       are certainly pleased to see it succeed in this area, and in
       particular the fact that it is being used at the Air Force
       Academy for teaching Ada, having been chosen by them as the
       best product for the task, is pleasing. However, there is fierce
       competition from Object Ada, and my view is that having two
       low cost options for teaching is a good thing. Two is a magic
       number when it comes to competition, and the competition to
       provide the best possible environment for students learning
       Ada will definitely benefit those students. I should say in
       this area that we are planning to distribute the next version
       of the NT version of GNAT with AdaGIDE, the IDE written by
       Martin Carlisle at the Air Force Academy. This is GPL'ed code
       written in Ada that provides a very nice environment for 
       student use (and may well be found useful by advanced users
       as well).

    2. Commercial production use. This was of course very definitely NOT
       part of the reason for DoD's support. Although the DoD had poured
       a lot of money early on in Ada 83 days into the support of 
       production compilers (look up how much was spent on ALS and ALSN
       for example), it was clearly not viable for the DoD to support
       the GNAT project on these grounds.

       However, the open nature of the development, and the use of the free
       software approach, meant that there was no way to restrict this
       possibility. Indeed, we have always taken the view that the needs of
       the academic community are not so different from the commercial
       production community. Both need high quality, full language
       compilers. Both can benefit from the availability of sources, and
       free redistribution.

       We always intended to produce a compiler that would compete
       effectively with proprietary compilers, because we saw this as
       the mechanism for continued support and development of GNAT. At
       first, we assumed that Cygnus would take on this task. However,
       Cygnus had no interest in Ada, and we later realized that we 
       would have to form our own company to follow this idea. Actually
       I think that works out well. Having ACT be 100% dedicated to
       Ada and Ada related products is more effective for the Ada 
       community.

So, I hope that give a clear background on the history of the GNAT
funding and project. Certainly the GNAT contract was one of the more
remarkable contracts signed by the government (it includes the full
text of the GPL, and, as I mentioned, explicitly requires the copyright
to be assigned to the FSF). The credit for making this happen belongs
to Chris Anderson, and her imaginative and flexible support staff at
Eglin Air Force base. A lot of tricky mazes had to be threaded to
get GNAT to happen, and Chris managed to find her way through them
where I am sure many others would have failed.

Robert Dewar
Ada Core Technologies