Re: Huge data files in Debian
On Sat, Jul 18, 2015 at 10:36:56AM +0200, Ole Streicher wrote:
> >> ? not sure if I would actually like to be "The Speaker" ;)
> > Why not. Your mail contained someconstructove details that could
> > kickstart a discussion. There is not more needed in a BoF.
> I'd support this -- we could also discuss the "distributed filesystem
> approach" there.
> Could this happen between Friday (DebCamp) and Tuesday?
> I have to leave on Wednesday...
I'm involved in scheduling. I hope to remember this ...
> For the specified package (astrometry.net), I have been looking into the
> content of the packages, and I think I can make a compromise here: The
> tables mainly contain star positions, and the tables range sizes range
> from brightest 1000 stars (for the smallest table) to ~200 million
> stars, roughly doubling with each table. Already the first 12 tables are
> quite useful for many applications (not for my instrument, however :-( ),
> covering 1.45 million stars, and have a size of 114 MB; I think this
> is acceptable for a Debian package.
Sounds acceptable, yes.
> However, there may be a licensing issue: The data are officially under
> GPL-2+; but this is sort-of impossible: GPL requires to have the
> "sources" available (and defines source as "the format that a human
> prefers to edit" or so). There is no such "source" for these files: they
> come from an survey (2MASS) and the final "source" are the positions of
> the stars on the sky. These positions are obviously not editable by
> humans yet :-) And while the process of generating the star catalogs is
> somehow documented, I doubt that we can or should reproduce the catalogs
> from the original exposures in Debian -- this would require disk space,
> computing power and man power which we obviously don't have.
May be ftpmaster might give some input here - if we are lucky one
ftpmaster will join the BoF.
> And the files have a quite straight-forward, documented structure so
> that anyone could patch them if needed.
> So, I would take these files (despite the fact that they are *generated*
> by another program) as source files. Maybe we could discuss this as well
> at the Debconf?
> At some point we should think about how to get this in
> our Social Contract.
No idea whether Social Contract is the proper place for this.