Position on large GPLed programs

[Matt Zimmerman]

On Sat, Sep 08, 2007 at 01:17:27PM +1200, Aaron Whitehouse wrote:
> What is the current position on big, GPLed programs? Do we just
> package anything GPL and then modify our approach when/if mirrors
> start to complain?

I'm a bit confused by the question.  What does the license have to do with
it?  A big, MIT-licensed program would seem to raise exactly the same
issues.

> To me it would make sense to have a consistent approach. I understand
> that the 700MB that are highest-rated in the popcon are put on the
> CDs. That seems like a good analogy to this issue.

That's not exactly how it works; popcon statistics are actually dominated by
the choice of default software, not the other way around.  While we do
consider relative popularity when selecting programs for the default
install, popcon is unfortunately not very useful for this at the moment.

> 1) I think that we should indiscriminately package anything that has
> an adequate license;
> 2) If mirrors start to complain, we should implement a new system so
> that mirrors can choose how much they are prepared to mirror.
> 
> The new approach could be tied into the
> https://launchpad.net/ubuntu/+archivemirrors
> system and some push-mirror idea:
> https://blueprints.launchpad.net/ubuntu/+spec/push-package-mirroring
> 
> If a mirror then chooses to only mirror 5GB of data, the Launchpad
> Mirror Manager could tell the mirrors the most-used 5GB of files to
> download, as rated by the popularity contest. If some push-mirroring
> system was in place, then the mirrors would be asking Launchpad Mirror
> Manager what they should be downloading anyway, so it shouldn't be
> overly difficult to put caps in place if they are requested. The only
> difficulty would be modifying the package manager to check other
> mirrors if the package that the user wants is not available on their
> preferred mirror. I suppose that a system like we currently use for
> security updates could be used.
> 
> What are people's thoughts?

Partial mirrors aren't very practical at the moment, and would require
substantial changes to the package management infrastructure, installer,
etc. in order to get it right.

We should not allow size to prevent good free software from becoming part of
Ubuntu, though we should of course make an effort to ensure that software is
not unnecessarily large.  For example, there was a thread some time ago
which measured whether PNG images could be compressed more efficiently using
pngcrush and similar tools.  In something like a game, I imagine that
graphical data represents a large portion of the size, and we should be sure
that the 250M is justified.

-- 
 - mdz