[Debian-med-packaging] Seeking Advice on 2.7G Package Data

[Olivier Sallou]

Le 10/8/12 1:58 PM, Laszlo Kajan a écrit :
> Dear Team, Andreas, Steffen!
>
> Our lab has a new free sub-cellular localization prediction method [1].
>
> We would like to package it, and it is almost done. The tool depends on an (arch indep) database that is 2.7GB (compressed, and it is used
> compressed). The data is (or will soon be) available as a tar.gz via FTP. The question is:
>
> * What to do with the data? - how to make sure it's available for the prediction method after it is installed?
>
>   1: We tried to create a 2.7GB loctree2-data package out of the data, and make loctree2 depend on it. Creating the package went well, but apt
> has problems with the size, it look like some bug in the stable version. This is not my preferred solution.
I think it should not Depends but Suggests it to avoid automatic
download. User may have the data, or use his own database.
A Recommends will make the download depending on configurations, so I
would keep a Suggests relationship.

Furthermore it can be painful depending on available bandwidth.
>
>   2: Create a loctree2-data-installer package that downloads the large data upon installation, flashplugin-installer style. I am worried this
> may be problematic for automatic testers (piuparts) at Debian, because of the large data it moves. Do I need to worry about this? Also I don't
> know how this behaves with interrupted downloads (continuing the download should be supported).
>
>   3: Have the executable download the data, or tell the user to download the data, when it is run and the data is not available, or outdated. My
> worry here is that this makes system-wide installation more complicated. The installation would be done by an admin, but the large data would be
> pulled in by an unprivileged user most likely, who can not install it into /usr/share/loctree2-data. The admin would have to be warned by the
> user that installation of the package is not enough.
I would suggest to ask user to download the database (in README.Debian
and in man page) manually to use the tool.
When running the tool, if database is not present, tool could ask to
download the db.

Olivier

> Please advise, and thanks in advance.
>
> Best regards,
>
> Laszlo
>
> [1] http://www.ncbi.nlm.nih.gov/pubmed?term=22962467
>
> _______________________________________________
> Debian-med-packaging mailing list
> Debian-med-packaging at lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/debian-med-packaging

-- 
Olivier Sallou
IRISA / University of Rennes 1
Campus de Beaulieu, 35000 RENNES - FRANCE
Tel: 02.99.84.71.95

gpg key id: 4096R/326D8438  (keyring.debian.org)
Key fingerprint = 5FB4 6F83 D3B9 5204 6335  D26D 78DC 68DB 326D 8438