Re: General Resolution: Interpretation of DFSG on Artificial Intelligence (AI) Models

Thanks a lot Mo for your work on this!

I have two quick comments:

  - In 2000, Debian's role in the relicensing of the Qt library was well
    recoginsed.  There are battles we can win.

    https://web.archive.org/web/20180324223759/http://www.linuxplanet.com/linuxplanet/reports/2269/1/
    (Link from https://en.wikipedia.org/wiki/Q_Public_License#cite_note-3)

  - Of course, chances will by higher if more people and projects join
    us.  One main argument against Mo's proposal is the cost difficulty
    of hosting the raw data.  For the models that are in Debian at the
    moment in Trixie, it is hard to think that the problem will not be
    solved if it really matters to the Free Software commmunity in
    general.

  - Even for retraining, I would be interested to hear a comparison
    between the environmental footprint of one Debian release, including
    all the continuous integration tests including mass rebuilds,
    autopgktests, reproducible build tests etc, plus the distribution via
    mirrors and CDNs, a couple of Debconfs, etc, compared with one or few
    retainings of all the models we actually ship, to check
    reproducibility of the output or at least the user experience…

  - Amendments or counter proposals that contain an exemption for time
    (releases, years…), either broadly or limited only to software
    that is already in Trixie, etc., are cheap to write, accept and act
    on.

Sorry if it has been already  discussed.  I will not have time to read
the whole thread (summaries welcome).

Have a nice week-end!

Charles