Re-licensing email

[Matt Mackall]

On Thu, 2009-10-15 at 12:52 +0100, Frank Kingswood wrote:
> Matt Mackall wrote:
> > On Wed, 2009-10-14 at 08:33 +0100, Frank Kingswood wrote:
> >> I got a surprising email this morning regarding relicensing Mercurial under 
> >> GPLv2+. There was (apart from the Mercurial-must-follow-Eclipse-license 
> >> thread) not any recent discussion on the mailing list about this.
> > 
> > This follows from the cvs2svn license discussion.
> > 
> > I took the issue up with our lawyers at the Software Freedom Law Center,
> > and had a fairly lengthy back and forth about all the possible loopholes
> > and counterarguments.
> 
> Just to stir some more into the mix, do you have any comments on this GPLv2 
> story?
> 
> http://www.theregister.co.uk/2009/10/15/black_duck_gpl_web_conference_copenhaver_radcliffe/

I'm pretty sure I'm not a copyright lawyer, or even a lawyer at all, but
I'll share my thoughts anyway

First off, it bears mentioning that the GPLv2 has an incredibly good
track record. -Lots- of large companies with significant legal resources
have (mostly inadvertently) violated the GPL, and just about all of them
that have been called on it have eventually caved and agreed out of
court to come into compliance. In the few cases where the GPL has been
brought to court, it has unambiguously prevailed, and in multiple
jurisdictions. And that's despite the free software community having
comparitively limited legal resources. You'd be hard-pressed to find any
non-trivial license, free or otherwise, that had a comparable batting
average.

The point about defining distribution is probably valid, and GPLv3 makes
a point of clearing this up. I'm not particularly concerned about this
point, as I think a court can arrive at a reasonable result about
whether some action is or is not distribution in just about any case
it's likely to hear. If you claim that company X must also distribute
source code for Fedora because they downloaded a copy of it for
themselves with BitTorrent (and were therefore technically
simultaneously uploading it), I don't think a court will be impressed,
especially when they hear the source is readily available from Red Hat
anyway.

The point about "derived works" entirely misses the point though.
Copyright law defines several boundaries. One of them is nebulously
defined in terms of "derived works". The reason GPLv2 uses this language
is to extend its boundary to exactly match the boundary defined in
copyright law. In other words, copyright law says 'works in set X are
covered' and the GPL says 'that same set X must follow the rules'. We
don't actually know what the set X is or even how it's defined (because
the law is intentionally nebulous and there isn't much precedent for
software), we just know that it's the same set in both cases.

-- 
http://selenic.com : development and support for Mercurial and Linux