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

Bug#481491: debian-policy: please add LPPL v1.3a to Policy



Russ Allbery <rra@debian.org> writes:
> Hilmar Preusse <hille42@web.de> writes:

>> In the moment we have the situation that we are requested to add the
>> license to all TL packages, instead of just adding it to one and put
>> only referers into the others (#473216). This gives us 1.8 MB of
>> license files, which could be saved of the LPPL would be in base-files.

> More relevantly than the total size, IMO (1.8MB isn't really very much)
> is that according to popcon, one-seventh of our systems have at least
> texlive-base installed.  If every texlive-base installation would
> benefit from having LPPL in common-licenses and most installations
> involve more than one package with the LPPL, that looks like a fairly
> reasonable case for common-licenses to me.

> It's not as strong of a case as the Apache 2.0 license had (nearly half
> of our popcon-reporting systems have at least apache2.2-common
> installed), but one-seventh is still a lot of systems.

> Manoj's suggested guide was 5% of the binary packages; that's probably a
> higher bar, and I'm not sure if TL meets that one; that would be about
> 4000 binary packages using the license, and we added the Apache license
> based on only about 250.  I wonder if something like 10% of
> popcon-reporting systems having at least two packages using that license
> installed would be a better metric.

I recently wrote a script to get us better answers for questions like this
and ran it across the entire archive.  Out of 29,740 binary packages in
the Lintian laboratory, 297 used the LaTeX Project Public License 1.3c
(which I checked for because it appears to be the most recent version).

Given that, I have several concerns with including this in
common-licenses.  First, we generally lean against including licenses
there because of all the advantages of having the license directly with
the package that it covers.  The main reason to put a license there is to
save space on users' systems and in the archive from having tons of copies
of the license.  One can particularly see this with the GPL, which is
referred to by 19,893 packages in Debian.  The least-used license family
currently in common-licenses is the GFDL, which is used by 875 binary
packages; this version of the LPPL is used by only a third of that.

Now, I did look for that particular version, and it's possible that the
numbers will go up if I do a broader search.  Do you think that's likely?
Are there a lot of packages under 1.3a (which this bug originally
referenced) or other versions of this license?

That also raises another worry, though, which is that apparently two new
versions of this license have come out since this bug was originally
filed two years ago.  For licenses, that's quite a bit of change, and if
we include it in common-licenses, we'll probably have to keep including
new versions as they're released and people update their software to use
them.  That means a lot of the benefit to common-licenses would be lost.
Is this license stabilizing, or is it likely to keep changing?

Given these facts, I'm personally currently leaning against including this
license in common-licenses, but I'm still open to being convinced,
particularly if you think that a broader search for more versions of the
license will turn up another significant cluster of packages.

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



Reply to: