Question on OSD #5

[Chris Travers]

On Nov 27, 2007 4:45 PM, Matthew Flaschen <matthew.flaschen at gatech.edu> wrote:
> Chris Travers wrote:
> > Hence they could merely use GPL'd libraries and argue that the use of
> > those libraries does not constitute derivation.   Suppose you do
> > challenge them?  They might win (and the linking == derivation
> > argument falls for everyone)
>
> No.  It falls for that particular instance.  Court decisions only apply
> to the relevant facts (and even then they get overturned, as with
> software patents, which were originally illegal and became legal without
> any relevant change in law).  Not every instance of linking is the same.

OK, I could have been more clear.  In theory you are correct, but in
practice the line is a little more muddled (IANAL, etc).

The FSF's licensing FAQ's seem to suggest that linking is equivalent,
or perhaps sufficient to prove derivation.  Any court judgement to the
contrary is likely to make it more difficult to push this position
categorically.  This means that it is no longer a new question before
the court and one has to convince them why the previous ruling does
not apply to the facts in the next case.  In short, the argument that
linking always implies derivation falls for everyone because now there
is a counter example to point to.

This means you have to show either that the first court was
fundamentally wrong and things need to be corrected (very hard where
statutory law is concerned)
or
That the facts in your case require a different sort of analysis.  In
short you have to show why you should be treated differently.

In short it would seem to result in a far greater difficulty of
showing that a given linking work was a derivative.

>
> > Governments are a sort of special case.  They can also rewrite the
> > rules in the middle in order to protect their interests.  In short it
> > may not be useful to suggest what governments may or may not be able
> > to do under the GPL.
>
> I agree to an extent.  Furthermore, given that the NSA /is/ one of the
> most secret agencies in the world, it's not like we could catch them
> violating the law.

There are all kinds of situations where they could get caught.
However, my points also would apply to any similar situation by any
other branch of government.

Hence it is not useful to speculate what someone can or cannot do when
they are writing the rules.

>
> You've got to be more specific.  There are half a dozen /types/ of
> patent retaliation clauses.  And we were really just talking about
> patent licenses, not retaliation clauses.
>
Ok.  Here is the classic example from the days when OpenAFS was
released under the IPL.  It applies to any piece of critical
infrastructure software and AFAICS applies to the ASL 2.0 and the GPL
v3.

Suppose there is software package X which includes a patent license
which is revoked when someone sues over patent infringement for any
patent as relates to that software.

Suppose companies A and B both rely heavily on that software in such a
way that migration from it is not something which can be done quickly
or without prohibitive cost.

Company A has a patent that might be useful to the software, an files
a patent application for it.  However, they do not want to contribute
the patent to the software an make no contributions to that effect.

Company B reads Company A's patent application, implements the feature
and gets the contribution accepted.  This could furthermore solely be
done as an attempt to undermine company A as a competitor.

Company A cannot sue anyone (even company B) over the essential theft
of the patent unless and until they migrate away from the software
package X.

We can probably agree that Company B's behavior would be unethical and
that the project should probably remove the questionable contributions
when they are aware of them.  So how do we create a patent retaliation
clause which does not allow this sort of abuse?

Best Wishes,
Chris Travers