Re: [libiberty copyright assignment audit] cp-demangle.c status

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Robert Dewar]

> If on the other hand, the file claims that it is a separate program, and 
> does not claim that it is part of GCC, then there is some question as to 
> whether it is in fact part of GCC or not.  (Normally I would say 
> that being distributed in the tarball for X would establish something 
> as being 'part of X', but there are numerous examples explicitly to the 
> contrary in the GCC tarballs.)  This status affects the impact of 
> the copyright assignment statements.  :-)  Gross, eh?

I disagree, the statement in a file cannot be authenticated easily. A
copyright assignment is a legal document that assigns certain items. What
items are covered is a matter of description in this contract.

The issue is wheter something *is* part of GCC, not whether it claims to be
if the copyright assignment assigns any code that is part of GCC.

I think it unlikely that the statement in the file would have very much
force (due at least in part fo the issue of authentication).

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Kai Henningsen]

neroden@twcny.rr.com (Nathanael Nerode)  wrote on 04.08.03 in <20030804204438.GA5928@twcny.rr.com>:

> Robert Dewar said:
> >Note that the copyright notice has nothing to do with whether or not it
> >is assigned.
> >The actual copyright status of any file is determined independently of
> >any
> >notice in the file itself.
>
> While I what you mean, this isn't actually quite correct, as I
> discovered.  The issue lies in copyright assignments which assign "all
> changes made to GCC..." or some such.  In this case the question of
> whether the file is part of GCC becomes paramount.
>
> If the file itself claims to be part of GCC, as well as being
> distributed with GCC, I think that's quite sufficient to establish that
> the file is 'part of GCC'.

No, I think (IANAL) the point here is subtly different. If the file is  
part of gcc but claims to be omething different, then someone who has an  
assignment on file that covers gcc but not the other thing, and who  
contributes to it, might not realize that the file actually is part of  
gcc, and thus might behave under the assumption that what (s)he does is  
not covered by that assignment - for example, (s)he might contribute code  
that his company has (say) published under the GPL, but is not willing to  
assign to the FSF.

That is exactly where the appearance of things can become more important  
than the truth of things, when it leads people to do stuff they otherwise  
would not have done.

So it actually matters that Zack says he assumed it was covered.

MfG Kai

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Kai Henningsen]

dewar@gnat.com (Robert Dewar)  wrote on 04.08.03 in <20030804201157.BC0BBF2D86@nile.gnat.com>:

> > My only change to cp-demangle was to have it include sys/types.h  --
> > totally insignificant from a copyright standpoint.
>
> Hmm, I am not sure I agree. From a copyright point of view, it may be the
> case that a #include is entirely equivalent to including the entire text of
> what is included (see GEAC vs Grace, Newark district court, unfortunately

So?

The act of writing the #include would still amount to no more than  
copy&paste under that assumption. If anything, that would make the authors  
of sys/types.h relevant as contributors, not the person who caused the  
include. Said person didn't write the *contents* of sys/types, after all.  
He has no IP rights whatsoever on those contents.

MfG Kai

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Robert Dewar]

> > The FSF's lawyers should probably decide what *their* opinion on this 
> > topic is, and what they consider sufficient evidence that the copyrights
> > were assigned, disclaimed, or nonexistent.  (I wouldn't be surprised if 
> > they decided that they didn't need anything additional, on the grounds that 
> > the people involved all agree that the FSF holds sole copyright -- but 
> > I don't actually know what they would decide in practice.)
> 
> Exactly.  Our legal opinions don't really matter much here.

Yes indeed, the law is not what sensible technical people think it should be :-)
So the sensible technical opinions of such people are not always relevant!

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Joe Buck]

In the end, it is up to the FSF to decide what legal risks it wishes
to take, to address hypothetical concerns such as those raised by Robert
Dewar (that a statement in a file might be false).  In tricky cases,
RMS will consult the FSF's lawyer.

I think that the way to handle this is simply the way Nathanael has
been doing it: do an audit, and for any anomolies try to ascertain the
history of the file.  If the FSF is satisfied that it owns the file,
it can direct us to modify comments in the file appropriately.


On Mon, Aug 04, 2003 at 06:49:12PM -0400, Nathanael Nerode wrote:
> What authentication problem?  This is the version of the file 
> distributed by the FSF and read by people modifying it.  See above for 
> why I think it constitutes evidence (when it's present).
> 
> People who modified the file while it did not contain such a statement,
> could reasonably have been under the impression that it was not part of 
> GCC and was not affected by the copyright assignment.  Similarly, the 
> FSF could reasonably have been under the same impression (they weren't 
> distributing anything making definite claims to the contrary).  Hence, 
> there's an ambiguity.  
> 
> I believe it would probably be concluded that the file was, in fact, 
> "part of GCC", but having the statement should provide more evidence 
> that it was, for the purposes of the copyright assignments.

I think that this is good enough, but the decision-maker here should be
RMS, since the FSF will be taking any legal risks if mistakes are made.

> The FSF's lawyers should probably decide what *their* opinion on this 
> topic is, and what they consider sufficient evidence that the copyrights
> were assigned, disclaimed, or nonexistent.  (I wouldn't be surprised if 
> they decided that they didn't need anything additional, on the grounds that 
> the people involved all agree that the FSF holds sole copyright -- but 
> I don't actually know what they would decide in practice.)

Exactly.  Our legal opinions don't really matter much here.

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Nathanael Nerode]

I said:
> If on the other hand, the file claims that it is a separate program, and 
> does not claim that it is part of GCC, then there is some question as to 
> whether it is in fact part of GCC or not.  (Normally I would say 
> that being distributed in the tarball for X would establish something 
> as being 'part of X', but there are numerous examples explicitly to the 
> contrary in the GCC tarballs.)  This status affects the impact of 
> the copyright assignment statements.  :-)  Gross, eh?

Dewar said:
>I disagree, the statement in a file cannot be authenticated easily. A
>copyright assignment is a legal document that assigns certain items. 
>What items are covered is a matter of description in this contract.
Yep.  The description raises a question of fact.

>The issue is wheter something *is* part of GCC, not whether it claims 
>to be if the copyright assignment assigns any code that is part of GCC.
Yes, certainly.  All kinds of evidence could be brought up regarding 
that question (is it part of GCC?) if it ever went to court.

I believe, however, that having a file *distributed by the FSF* (one 
of the parties to the copyright assignments) in the "GCC" tarball, containing 
a statement that the file was "part of GCC"; such that anyone modifying 
the file would presumably have to read that statement when modifying it, 
so that they (the other parties to the copyright assignments) would 
implicitly be accepting the claim that it was part of GCC; is 
strong evidence that the file is, in fact, part of GCC; or at least that
it was considered to be such by the parties to the copyright 
assignment, which is important in the context of the copyright 
assignment.

(Whew.)

>I think it unlikely that the statement in the file would have very much
>force (due at least in part fo the issue of authentication).

What authentication problem?  This is the version of the file 
distributed by the FSF and read by people modifying it.  See above for 
why I think it constitutes evidence (when it's present).

People who modified the file while it did not contain such a statement,
could reasonably have been under the impression that it was not part of 
GCC and was not affected by the copyright assignment.  Similarly, the 
FSF could reasonably have been under the same impression (they weren't 
distributing anything making definite claims to the contrary).  Hence, 
there's an ambiguity.  

I believe it would probably be concluded that the file was, in fact, 
"part of GCC", but having the statement should provide more evidence 
that it was, for the purposes of the copyright assignments.

Zack's statement that *he* believed the file was covered by his 
copyright assignment should be evidence that it was covered as well. 
:-)

(I have, incidentally, assumed that everyone involved knew that "GCC" 
and "GNU CC" referred to the same thing, since I've never encountered 
anyone who didn't.)

Disclaimer:  This is not a legal opinion.  (Heh... it's a 'factual 
opinion', I guess...)

The FSF's lawyers should probably decide what *their* opinion on this 
topic is, and what they consider sufficient evidence that the copyrights
were assigned, disclaimed, or nonexistent.  (I wouldn't be surprised if 
they decided that they didn't need anything additional, on the grounds that 
the people involved all agree that the FSF holds sole copyright -- but 
I don't actually know what they would decide in practice.)

-- 
Nathanael Nerode  <neroden at gcc.gnu.org>
http://home.twcny.rr.com/nerode/neroden/fdl.html

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Kaveh R. Ghazi]

 > Three people other than Alex Samuel contributed potentially
 > coprightable work to the file prior to the copyright change: Kaveh
 > Ghazi, Zack Weinberg, and Jeff Law.

FYI in additional to my GCC assignment, I have assigned past and
future changes to libiberty.  So that should take care of it.

Looking at my changes, those that I made prior to version 1.23 of the
file were all small maintenance stuff, cleanups, constification, etc.
No functional changes, so they are probably not copyrightable anyway.

http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libiberty/cp-demangle.c.diff?r1=1.2&r2=1.3
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libiberty/cp-demangle.c.diff?r1=1.3&r2=1.4
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/libiberty/cp-demangle.c.diff?r1=1.10&r2=1.11

--
Kaveh R. Ghazi			ghazi@caip.rutgers.edu

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Matt Kraai]

Robert Dewar wrote:
> Hmm, I am not sure I agree. From a copyright point of view, it may be the case
> that a #include is entirely equivalent to including the entire text of
> what is included (see GEAC vs Grace, Newark district court, unfortunately
> no written decision from Appelate court here, but a ruling in GEAC's favor
> [despite Jury verdict to the contrary] that says not only this, but also says
> that just calling a routine is equivalent from a copyright point of view to
> including its entire text.

You can find the appellate court decision at

 http://www.ca3.uscourts.gov/opinarch/002772.txt

-- 
Matt

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Nathanael Nerode]

Robert Dewar said:
>Note that the copyright notice has nothing to do with whether or not it 
>is assigned.
>The actual copyright status of any file is determined independently of 
>any
>notice in the file itself.

While I what you mean, this isn't actually quite correct, as I 
discovered.  The issue lies in copyright assignments which assign "all 
changes made to GCC..." or some such.  In this case the question of 
whether the file is part of GCC becomes paramount.  

If the file itself claims to be part of GCC, as well as being 
distributed with GCC, I think that's quite sufficient to establish that 
the file is 'part of GCC'.

If on the other hand, the file claims that it is a separate program, and 
does not claim that it is part of GCC, then there is some question as to 
whether it is in fact part of GCC or not.  (Normally I would say 
that being distributed in the tarball for X would establish something 
as being 'part of X', but there are numerous examples explicitly to the 
contrary in the GCC tarballs.)  This status affects the impact of 
the copyright assignment statements.  :-)  Gross, eh?

It's the (former) absence of the 'part of GNU CC' phrase in the 
file which is the actual issue for the copyright assignment status, 
rather than the (former) claim that it was copyright CodeSourcery.

>It is of course important that the notice in the file reflect
>the actual status.
Indeed. :-/

-- 
Nathanael Nerode  <neroden at gcc.gnu.org>
http://home.twcny.rr.com/nerode/neroden/fdl.html

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Robert Dewar]

> I note that the change log for revision 1.23 is
> 
> revision 1.23
> date: 2000/09/08 17:50:54;  author: samuel;  state: Exp;  lines: +3 -1
>         * cp-demangle.c: Fix copyright banner.
> 
> which suggests that the initial copyright notice was a simple error
> and it was intended to be assigned to the FSF from the beginning.

Note that the copyright notice has nothing to do with whether or not it is assigned.
The actual copyright status of any file is determined independently of any
notice in the file itself.

It is of course important that the notice in the file reflect
the actual status.

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Zack Weinberg]

"Zack Weinberg" <zack@codesourcery.com> writes:

>
> Whatever changes I made to that file way back then (I don't remember
> what they were) were made on the assumption that they were covered by
> my blanket copyright assignment for GCC.  I doubt I even noticed that
> the file was listed as copyright CodeSourcery.

I looked back through cvs log - the only change of mine in that file
prior to revision 1.23, which is when it was relabeled copyright FSF,
is

revision 1.7
date: 2000/06/09 18:55:36;  author: zack;  state: Exp;  lines: +23 -7
branches:  1.7.2;
        * cp-demangle.c (demangle_operator_name): Add spaces before
        names beginning with a letter: delete, delete[], new, new[],
        sizeof.
        (demangle_special_name): Handle TF <type> and TJ <type>.

The "add spaces before" bit is trivial, the obvious way to fix a bug
(printing of _Znwj as "operatornew(unsigned int)").  The TF/TJ special
name bit is bigger and not quite so trivial, but I would still call it
small enough not to worry about.

I note that the change log for revision 1.23 is

revision 1.23
date: 2000/09/08 17:50:54;  author: samuel;  state: Exp;  lines: +3 -1
        * cp-demangle.c: Fix copyright banner.

which suggests that the initial copyright notice was a simple error
and it was intended to be assigned to the FSF from the beginning.

zw

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Robert Dewar]

> My only change to cp-demangle was to have it include sys/types.h  -- totally
> insignificant from a copyright standpoint.

Hmm, I am not sure I agree. From a copyright point of view, it may be the case
that a #include is entirely equivalent to including the entire text of
what is included (see GEAC vs Grace, Newark district court, unfortunately
no written decision from Appelate court here, but a ruling in GEAC's favor
[despite Jury verdict to the contrary] that says not only this, but also says
that just calling a routine is equivalent from a copyright point of view to
including its entire text.

No point in debating this specific point, but my general point is that it is
risky to make any legal judgments about copyrights even if what you say is
obviously reasonable. best to make sure all necessary documentation is in
place so no issue ever arises.

Re: [libiberty copyright assignment audit] cp-demangle.c status

[law]

In message <87n0epryed.fsf@egil.codesourcery.com>, "Zack Weinberg" writes:
 >neroden@twcny.rr.com (Nathanael Nerode) writes:
 >...
 >> The change in the copyright statement probably corresponds to a copyright
 >> assignment by CodeSourcery.  If it does, then presumably all of Alex Samuel
 >'s
 >> previous work is assigned and OK.  (It's also possible that the CodeSourcer
 >y
 >> copyright was an outright error, in which case I'm not sure what to do...)
 >>
 >> Three people other than Alex Samuel contributed potentially coprightable wo
 >rk
 >> to the file prior to the copyright change:  Kaveh Ghazi, Zack Weinberg, and
 >> Jeff Law.  
 >>
 >> I doubt that contributions to something which was not then
 >> marked as being part of GCC are technically covered under GCC copyright
 >> assignments or disclaimers.  Jeff Law was at Cygnus at the time, so his
 >> contribution may (or may not) have been covered by a later Cygnus assignmen
 >t;
 >> the others appear to have been private citizens.  ;-)  
 >>
 >> Since all four people are current GCC maintainers, I hope that the issue
 >> of copyright assignments on this file can be cleared up reasonably quickly.
 >
 >Whatever changes I made to that file way back then (I don't remember
 >what they were) were made on the assumption that they were covered by
 >my blanket copyright assignment for GCC.  I doubt I even noticed that
 >the file was listed as copyright CodeSourcery.
My only change to cp-demangle was to have it include sys/types.h  -- totally
insignificant from a copyright standpoint.

Jeff

Re: [libiberty copyright assignment audit] cp-demangle.c status

[Zack Weinberg]

neroden@twcny.rr.com (Nathanael Nerode) writes:
...
> The change in the copyright statement probably corresponds to a copyright
> assignment by CodeSourcery.  If it does, then presumably all of Alex Samuel's
> previous work is assigned and OK.  (It's also possible that the CodeSourcery
> copyright was an outright error, in which case I'm not sure what to do...)
>
> Three people other than Alex Samuel contributed potentially coprightable work
> to the file prior to the copyright change:  Kaveh Ghazi, Zack Weinberg, and
> Jeff Law.  
>
> I doubt that contributions to something which was not then
> marked as being part of GCC are technically covered under GCC copyright
> assignments or disclaimers.  Jeff Law was at Cygnus at the time, so his
> contribution may (or may not) have been covered by a later Cygnus assignment;
> the others appear to have been private citizens.  ;-)  
>
> Since all four people are current GCC maintainers, I hope that the issue
> of copyright assignments on this file can be cleared up reasonably quickly.

Whatever changes I made to that file way back then (I don't remember
what they were) were made on the assumption that they were covered by
my blanket copyright assignment for GCC.  I doubt I even noticed that
the file was listed as copyright CodeSourcery.

zw

[libiberty copyright assignment audit] cp-demangle.c status

[Nathanael Nerode]

This is a copyright assignment 'audit' for libiberty/cp-demangle.c.

First, the data:
----------------

* Original file entered into CVS was written by Alex Samuel and copyright
CodeSourcery LLC.  Listed as 'This program' in copyright statement.

* The following people made modifications while it was listed under this
copyright.  They are all current GCC maintainers:
- samuel - Alex Samuel
  -- presumably originally under the same CodeSourcery copyright
- ghazi - Kaveh Ghazi
- phdm - Phillipe De Muyter 
  -- changes are minimal, all deletions, certainly not copyrightable
- zack - Zack Weinberg
- law - Jeff Law

The copyright header was changed by Alex Samuel (who presumably knew what
he was doing) in revision 1.23, to be copyright the FSF.
This also introduced the statement "This file is part of GNU CC".

Subsequent changes were made by (all current GCC maintainers):
- jason - Jason Merrill
- bryce - Bryce McKinlay
- samuel - Alex Samuel
- dj - DJ Delorie
- jimb - Jim Blandy
- zack - Zack Weinberg
- jakub - Jakub Jelinek
- ghazi - Kaveh Ghazi
- pme - Phil Edwards
- hjl - HJ Lu

Mark Mitchell added the libgcc exception to the copyright in revision 1.40.
Subsequent changes were made by (all but Carlo current GCC maintainers):
- pme - Phil Edwards
- bryce - Bryce McKinlay
- mmitchel - Mark Mitchell
- ghazi - Kaveh Ghazi
- jason - Jason Merrill
- danglin - Dave Anglin
- Carlo Wood
  -- single typo fix, not copyrightable

Now, the analysis:
------------------

Following the change in copyright statement to describe the file as
"Part of GNU CC", I believe that GCC copyright assignments and disclaimers
apply to it.  Since all the copyrightable edits were done by GCC maintainers,
I presume they were paperwork-complete for GCC/"GNU CC", so I'm not going to
worry about those.

The change in the copyright statement probably corresponds to a copyright
assignment by CodeSourcery.  If it does, then presumably all of Alex Samuel's
previous work is assigned and OK.  (It's also possible that the CodeSourcery
copyright was an outright error, in which case I'm not sure what to do...)

Three people other than Alex Samuel contributed potentially coprightable work
to the file prior to the copyright change:  Kaveh Ghazi, Zack Weinberg, and
Jeff Law.  

I doubt that contributions to something which was not then
marked as being part of GCC are technically covered under GCC copyright
assignments or disclaimers.  Jeff Law was at Cygnus at the time, so his
contribution may (or may not) have been covered by a later Cygnus assignment;
the others appear to have been private citizens.  ;-)  

Since all four people are current GCC maintainers, I hope that the issue
of copyright assignments on this file can be cleared up reasonably quickly.


--Nathanael


-- 
Nathanael Nerode  <neroden at gcc.gnu.org>
http://home.twcny.rr.com/nerode/neroden/fdl.html