[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: GPL-like numbers, and monetary value

David Prévot <taffit@debian.org> writes:
> Le 18/02/2012 07:16, Francesca Ciceri a écrit :

>> As usual, the issue is available on the publicity subversion repository,
>> even via HTTP:
>> http://anonscm.debian.org/viewvc/publicity/dpn/en/current/index.wml?view=co

> I'm wondering about John Sullivan' numbers in the GPL article: Russ,
> since you often check for such numbers in order to include licenses in
> /usr/share/common-licenses, can you please confirm the claim that “93%
> of Debian Squeeze's packages are released under licences in the GPL
> family (including the GPL, the Affero GPL and LGPL)”?

The tool that's being used for this purpose isn't designed for that
purpose, so those numbers should be taken with quite a bit of salt.  My
assumption is that the number is derived from adding:

AGPL 3                   54
GFDL (any)             1038
GPL (any)             24192
LGPL (any)             9326

together, which gives you 34,610 out of 36,940 packages, which looks very
high.  However, the tool is designed for a specific purpose in auditing
license usage for Policy and therefore does not classify packages into one
and only one category.  So you've just double-counted every package that
has both GPL and GFDL, or both GPL and LGPL, content.

Furthermore, the tool only looks for packages with any reference to those
licenses in the copyright file.  This has other problems.  For example,
most of my packages will show up, erroneously, as being under the GPL
because they mention that the libtool script and some other packaging
files is distributed under your choice of the GPL or the license of the
overall package.  This means the package has some GPL content, sort of,
but this isn't really what anyone thinks about when they think of the
package as being released under the GPL.  The packages are really under an
MIT license, or an Apache 2.0 license, depending on which package.

So, basically, these numbers are consistent with the output of
license-count, but license-count wasn't designed to do this analysis and I
think the conclusions drawn from it are dubious at best.

Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>

Reply to: