On 10.06.2010 21:45, Russ Allbery wrote:
I recently did a survey of both licenses already listed in common-licenses
and ones proposed for common-licenses using a Perl script that's now in
the debian-policy Git repository. The result was that the MPL version 1.1
was used by 654 binary packages in the archive.
This is by far the best claim of any of the proposed new licenses for
common-licenses, although it still falls short of the least-used license
already in common-licenses (the GFDL, used by 875 binary packages in some
variant or version) and certainly well short of the 5% of the archive
standard that Manoj proposed (which would be 1473 binary packages).
On the other criteria we've used in the past, popcon, the MPL 1.1 fares
relatively well, since Iceweasel references it and could replace its copy
with a link to common-licenses and is installed by 50% of the systems
reporting via popcon.
I'm not sure to what extent including something in common-licenses is an
approval. We included the GFDL because it was widely used and looked
likely to be more widely used, despite the fact that the project
definitely isn't very fond of it.
So I'm torn on this one, and the discussion also seemed divided. I'm
leaning mildly towards rejecting it, but only very mildly.
The common-licenses was done (IIRC) to save disk space, so to
use such criteria, I would count only packages with priority >=
standard, or a proof that most systems have the verbatim license
installed many times).
OTOH the disk usage is no more a big issue, so we could use
common-license as a pool of common and recommended licenses.
Personally I don't think policy should discuss so many licenses,
so, I would like:
- make clear and strong requirements for new licenses (e.g.
we should include only few licenses), or
- move the choice outside policy procedure (e.g. maintainer