Re: Do we need Debian(-Med) machines?
Le Sun, Aug 12, 2007 at 01:00:18AM +0200, Steffen Moeller a écrit :
> Dear ...,
> we met at the ISMB in Vienna where I learned a lot about your PowerPC-Cell
> platform and took the opportunity to introduce you to the Debian Linux
> Bioinformatics community. I asked for the possibility to grant access to such
> a machine to developers for the Debian community to promote that platform.
> We agreed that it might be mutually beneficial to such establish closer ties
> with the Debian community and, if you are still interested, I can only offer
> to help evaluating possible routes to implement these.
> There is a second not unrelated issue in which I perceive you as a potential
> ally. The developers from the Debian-Med community
> (http://www.debian.org/devel/debian-med/index.de.html) become increasingly
> aware that for their Bioinformatics support the next strategic move is the
> integration of public data with to the already available software. The
> frequent updates of the data paired with the considerable size of the data
> however prohibits their distribution via regular Debian channels and its
> mirrors. Furthermore, the indexing of data needs to be performed in strict
> synchronicity with the data and is considerably time consuming. To have such
> files available as Debian packages would solve these issues. May I bluntly
> ask if you would be prepared to help with a (non backuped) server for a
> community project that is addressing the provisioning of bioinformatics data
> packages for Debian Linux? Maybe on a Cell-based machine?
> Some side thoughts: Bioinformatics Debian may experience a further boost with
> the advent of the BOSC liveCD (http://open-bio.org/wiki/BOSC/liveCD) which
> will be based on Debian. If essential bioinforamtics data can be seemlessly
> collected into packages, then a general acceptance of those may be beneficial
> for the local evaluation of scripts prior to the submission to deep computing
> facilities or grid computing in general.
Indeed, the data indexing and redistributing is one of the steps we can
not do without external help, so your letter makes a lot of sense. Due
to its highly mirrored nature, it would definitely not be realistic to
claim gigaoctets in the Debian archive. I still fancy the idea of using
a port system such as Fink to generate Debian packages, using build files
officially distributed in Debian so that the whole system is
reproducible. This way, we could build the data packages on a Debian-Med
machine and provide them in parallel. I have read that indexing big
databases such as Genbank really takes time, so it would be an useful
service for facilities which have to be up and up-to-date 24/7.
Have a nice day,
Wako, Saitama, Japan