Copyright of the desktop configuration specification (careful, here be dragons)

[Jim Gettys]

There is another problem with the GFDL is the invariant clause that the
Debian document you provided a link to misses entirely:

The invariant clause can make documentation and/or a specification
useless in the case of a fork in the code (without rewriting the
documentation from scratch), due to bad interactions with trademark law.

Say invariant sections surround uses of a trademark, and these sections
represent key or large parts of the document; if the code and/or
specification needs to be forked (as happens occasionally, for good or
bad reasons), you may not be able to eliminate the possible confusing
use of the trademark, and therefore be in violation of Trademark law.
You could be taken to court over this one...

At times the documentation/specification may represent a good fraction
of the total effort in a system.  I remember spending many months
working on Xlib documentation/specification in the '80's and early
'90's; the code as written at that date was a few man years, so the
documentation/specification was 10's of percent of the total effort; if
you throw in the tech writer's time on top of my time, it may have been
as much as 40%.

So for all the reasons outlined by Debian in the link you provide below,
and the reason outlined above, the GFDL is useless, and subtly so in
ways that can be pernicious and not obvious on first examination (which
is how it came to be, and why Debian to struggled over it). The freedom
to fork code for your own purposes is a sometimes forgotten, but
essential freedom; losing all the work in the documentation and
specifications can be a *major* compromise of that freedom, whether you
come down on the free or open source side of the fence.  This problem
only occurred to me as we were cleaning up the use of trademarks in the
XFree86/X.org fork in the documentation.

It isn't clear I've seen a copyright I like for formal specifications
for free/open source software, though I suspect we could generate one;
IIRC the IETF has copyrights that might be a starting point.  The issue
I see pretty unique to specifications is ensuring that the spec not be
claimed to be the specification unless it really *is* the agreed to
specification of the organization blessing the spec.  In some cases, the
specification also serves as the documentation for code (and vice
versa); in other cases, it is an original work unto itself being
implemented by multiple code bases.

I'll ping Eben Moglen on this one, along with Michael Tiemann.  It is a
point of law that I think hasn't seen enough thought in open source/free
software, and one that will be increasingly common as we "grow up" and
have more and more formal specifications for our important systems and
protocols.  It would be good if the SFLC and OSI could tackle this issue
and come up with a single good copyright for documentation and
specifications, and avoid lots of copyright proliferation, and the
resulting headaches of bad interactions between these copyrights and the
copyrights used on the code itself.
				- Jim


On Wed, 2005-09-28 at 13:52 +0200, Philip Van Hoof wrote:
> On Tue, 2005-09-27 at 19:41 +0200, Philip Van Hoof wrote:
> 
> > I'm guessing the LGPL license or the GNU Free Documentation License are
> > interesting ones.
> 
> One person (in private) expressed his doubt about the GNU Free
> Documentation License. Mainly because Debian doesn't like this one:
> 
> http://people.debian.org/~srivasta/Position_Statement.html
> 
> Other opinions?
>