Packaging special packages
Hi Science packagers,
Some time ago, there was a small discussion between me and Julian Taylor
about the packaging of a special package [1], which was also forwarded
to this list.
However, I would like to re-start this discussion now and get some more
opinions since the problem may exist for a couple of scientific software
packages:
The European Southern Observatory runs one telescope (VLT) in Chile
which uses several "instruments" (camera, spectrograph etc.) to get the
data. The data processing for these instruments is very specific and is
done in so-called "pipelines", from which about 20 exist [2]. Their
structure is quite similar, so once the first pipeline is packaged, the
rest doesn't require much effort. The dependencies of the pipelines are
already packaged in Debian.
However, there is one critics, that was brought up by Julian: every
pipeline can be used only for one specific instrument on this unique
telescope. If one doesn't have observational data from the VLT for that
specific instrument, the pipeline is worthless. And usually all
observations are done by a specific request of a scientist to fullfill
his needs.
On the other hand, these data become freely available for everyone [3],
allowing (and ecouraging) the scientific re-use by the whole
community. Especially for scientists without direct access to the
telescope, this is an excellent opportunity for scientific work. I think
that this perfectly fits into the best goals of the "Debian ethics".
Typically, binary packages of the pipeline code would have a footstamp
of <~ 1MB. However, the pipelines are usually accompanied with a
calibration data set, whose size ranges from some 100 kB to ~100 MB. The
calibration files are needed to actually run the pipeline with some
scientific result. I think it would be in any case too much to put all
these data onto Debian mirrors, just for the few astronomers out there.
So, having a package that downloads and installs the calibration data
would be the best here, right? But this would make the packages no
longer self-contained. Would that be a legal problem for a Debian
package in main?
What do you think: is it worth to upload these "pipeline" packages to
Debian? Or is it better to keep them in some personal repository?
Best regards
Ole
[1] http://bugs.debian.org/709330
[2] http://www.eso.org/sci/software/pipelines/
[3] http://www.eso.org/UserPortal
Reply to: