Re: SGI GNAT Question? (Long)

[dewar@gnat.com]

In article <36E25778.C056829@chocolatesaltyballs.com>,
  root <white@chocolatesaltyballs.com> wrote:

Well a lot of confusion and misunderstanding here, so it
is a useful opportunity to clarify things:

> Note the use of the "if" at the start. Are you claiming
> literal ownership of GNAT? what do your fellow
> contributors ( from your mates(?) at NY to joe bloggs on
> the street who sent in a GPL bug fix to a GPL source )
> say to that? Are they they ALL in unison ?

The copyright for each part of the GNAT system is no
secret. You can find it in the appropriate unit in the
sources. There are basically the following copyrights
around:

1. Copyrights that have been assigned to the FSF by
NYU for the original work done at NYU (by the way almost
everyone on the NYU team now works full time for ACT!)
This was done by a normal FSF assignment. As always with
the assignment document, the original assignee retains
an unlimited license to do anything they want with the
program, but that license rests with NYU, and they are
unlikely to excercise it.

2. Copyrights that have been assigned to FSF by ACT for
follow on work. Obviously we could use our own copyright
for fixes and improvements to the basic GNAT modules, but
we choose not to, because it is our corporate policy that
all FSF copyrighted units continue to have the full FSF
copyright (see more discusssion below).

3. Copyrights that ACT holds. These are used for major
components developed after the NYU contract, some examples
are the SPITBOL routines in the GNAT library, and the JGNAT
backend. ACT is commited to releasing these components
under the GPL including future versions, but as I have
pointed out, this is a function of corporate policy, not
of the legal requirements of the GPL.

4. Copyrights held by third paries. Examples are the
tasking units which have an FSU copyright. These are
all covered by the GPL, and that is the basis on which
ACT uses these units. We intend to continue to provide
them in GPL'ed form, but of course the third party
copyright holders could excercise their rights to make
non-GPL'ed versions in the future.

> because if one of those contributors says NO, end of
> story.

Well it is worse than this if you think about it. What if
someone makes a fix that is not under the GPL, then you
have a deriviative work that cannot be distributed without
the explicit permission of both copyright holders (because
of course the GPL one one part does not permit you to
distribute other things not covered by the GPL). Note that
in this case the holder of the copyright on the non-GPL'ed
modification cannot distribute the combined work, since
the GPL prevents this.

What the FSF does, and what we do, is to accept changes
only if the copyright situation is dealt with
satisfactorily. There are various possibilities here

1. Small fixes are considered de minimis and
non-copyrightable. A usual standard has been that anything
up to ten lines is covered by this (this is the standard
that for example gcc has used). We prefer to treat even
small changes as in para 2 following where possible, but
certainly if someone points out that a word is wrongly
spelled in the documentation, they do not have an
enforcable copyright interest in the correct spelling :-)

2. Larger contributions must involve either an assignment
of copyright, or a waiver of copyright interest. At ACT,
we have a large file of such waivers on hand, and that is
a normal procedure for the maintenance of GPL'ed software.
It is just not feasible to have a single file in which many
different people have a legal copyright interest.

3. Public domain work of any kind is of course not a
problem. Essentially this is work for which the author
has in advance issued a waiver.

4. Completely separate units can be incorporated into GNAT
with the original author's copyright where appropriate.

> What would the original US government dept that sponsored
> the development of a "free" compiler say to that?

The contract with the US government had two requirements:

1. All units were to be released under the GPL or LGPL (we
later decided on the more permissive GNAT modified GPL for
the latter purpose).

2. All copyrights were to be assigned to the FSF

Most certainly the original US government dept that
sponsored GNAT has no control or say in what happens at
this stage. The contract was completed, and the conditions
of the contract were met to the satisfaction of the
government auditors.

If ACT or some other entity had decided to maintain GNAT
on a proprietary basis (the GPL makes this trickier but
by no means impossible), then that would have been too bad
but not a contract violation.

Really what I am stressing here is that neither the GPL
nor the expired government contract require that GNAT be
handled in a completely open manner (e.g. that public
releases continue to be made). This is a function of
clearly stated ACT policy.

I constantly run into people who think that ACT is somehow,
by someone, or something, *required* to make public
releases of GNAT. But this is definitely not the case,
it is something we are committed to do, because we feel
it is an important contribution to the Ada community. The
ACT policy here has been absolutely clear from the start,
and I will restate it again, so there can be no confusion:

"It is the policy of Ada Core Technologies, and ACT/EUrope,
that GNAT and its associated tool set will continue to be
distributed under the GPL (or GGPL) and that public
versions will be released from time to time that reflect
the complete state of the technology in open source form."

> If I or anyone who came by the GNAT source through the
> GPL modified it and released it to others it would have
> to be GPL as stated in section 4 and 5.

If you have a legitimate license (the GPL in this case),
you can certainly modify it and distribute the modification
in any form allowed by the GPL (your modification does not
have to be under the GPL, it can be under a more liberal
license, e.g. be in the public domain, that is up to you).

For ACT to incorporate this change into the mainstream
sources, the copyright issue would have to be resolved
as described above.

> What you are saying is that you never can by the source
> through the GPL therefore you are not covered by it.
> Interesting point.

No no, that's not quite right. The *copyright holder* does
not need a license to use their own stuff. They can do
anything they like. In the case of ACT, we definitely need
a license to use other people's copyrighted software:

  1. The components where the FSF holds the copyright
  2. The components where some third party (e.g. FSU) holds
     the copyright

and we rely on the GPL for this access. For units to which
ACT holds the copyright, we certainly need no license at
all!

> but it only takes one other contributor to say that their
> code fix which you incorporated was released under the
> GPL for you to be bound by it.

As I made clear above, we would never incorporate changes
without resolving the copyright issue as above. Note that
in the case of assignments, we would use the FSF assignment
form which is an unusual one, because it constrains the
recipient of the assignment to continue distribution under
the GPL.

Suppose for example, you right some GPL'ed unit, you then
assign the copyright in the normal manner to IBM. They then
make a proprietary version, which they hold the copyright
to, and refuse to license it, even to you. Sound unfair?
Yup, but this kind of assignment is not uncommon (most
authors have to execute such assignments to their
publishers for example).

The FSF assignment, which ACT would also use, has two
important extra conditions:

1. It constrains the assignee (FSF or ACT) to always
release future modifications and versions under the GPL.

2. It grants an unlimited license back to the assignor

> ALL versions of GNAT are GPL and they always will be,
> Richard Stallman probably has a quad of legal cash
> waiting for the first commercial enterprise to abuse the
> GPL, as that would set the precedent.

Richard Stallman would only have legal standing if we
did something inappropriate with an FSF copyrighted
component, and even there it would not be Richard Stallman,
but rather the FSF that had standing.

The GPL itself is a license which is a contract between
two parties. There are only two kinds of legal redress
available, violations of the contract under contract law,
which of course can only involve the parties to the
contract, and copyright violations, which can only
involve the copyright holder.

If FSF is neither a party to the contract, nor the
copyright holder, then they are not an interested party.
Richard Stallman might or might not give his opinion
on the appropriate interpretation of the GPL, and might
for example be an expert witness on one side or the other
in any legal proceeding, but would not otherwise have
standing.

I often find this is a confusion, people somehow think that
any instance of the GPL involves the FSF. This of course
is not the case. The GPL is just a license which anyone
is free to use in any situation they like.

> You can no more take GNAT back, that Larry Walls, Linus
> Torvalds and Richard Stallman could take back perl,
> linux, or emacs.

Well you are probably right from a practical point of view,
but the point is that nothing would legally prevent these
authors from trying to make a proprietary version of their
work. Surely they would fail in the attempt.


> > When you get a licensed product from Microsoft, you
> > know perfectly well that they still own the program and
> > can do anything they like with it.
>
> So you have to use Microsoft as an example to make
> yourself look good in the GPL world ?

That's an odd statement. This has nothing to do with
looking good, it has to do with legal requirements of
the GPL, which is what we are discussing.

The point here is that historically, lawyers for large
companies have often been very suspicious of the GPL,
because they think something funny is going on (indeed
this message which I am responding too makes an attempt
to reinforce this incorrect viewpoint).

Realizing that the GPL is just a license, no different
in fundamental legal structure form the licenses that
Microsoft grants is very helpful, because it makes large
companies see that there is nothing to be afraid of in
using GPL'ed software. We have spent a lot of time working
with large companies to alleviate concerns of this kind.

> > Well there is *nothing* unusual about the GPL in this
> > regard, it is simply a limited license giving the
> > recipient of the license certain limited rights to use
> > the copyrighted works.

> The terms copyleft

Actually this is not a technical term, but just a popular
one. It is probably unfortunate since it has tended to
create the impression that GPL'ed software is not
copyrighted, but rather exists in some strange legal
state different from copyright. Note that the GPL itself
(quite deliberately) does NOT use the term copyleft.

> and its simply a licence as a Rolls Royce and a Lada
> are cars.

Now *there* we can agree. The Rolls Royce and the Lada
are of course just cars, but they are cars you would like
to have if you have the choice.

The GPL is indeed a Rolls-Royce of licenses from the point
of view of the licensor, and helping companies to realize
that this is the case, and helping them to realize that
the GPL is highly desirable from their point of view is
what this is all about.

Robert Dewar
Ada Core Technologies

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own