[Bioperl-l] Packaging bioperl for Fedora

[Alex Lancaster]

On Mar 30, 2007, at 11:02 PM, Allen Day wrote:

[...]

>> If we're in agreement that the primary data sets and
>> libraries/applications for producing derivative data should not be
>> present in Fedora Extras, then it follows that the Bioperl classes
>> for manipulating these primary and derivative data should also not
>> be present in Fedora Extras as they are of little use without data
>> to manipulate.

Chris Fields wrote:

CF> I respectfully disagree.  BioPerl, to me, is a toolkit which helps
CF> accomplish certain tasks.  As with any toolkit, not all parts are
CF> required to do what one needs.  A good number of end-users use
CF> BioPerl for remote database queries
CF> (Bio::DB::GenBank/Taxonomy/etc), remote BLAST, seq analysis,
CF> alignment analysis, phylogenetic tree manipulation, etc, none of
CF> which require outside apps be installed.  For many a remote db is
CF> their primary source of data; not everybody sets up BioPerl for
CF> accessing local db records, running programs, etc (just the smart
CF> ones!).  As for outside apps, the docs are pretty explicit where
CF> certain outside resources (libxml2, expat, libgd) are needed for
CF> functionality.

CF> When we package up a new release we generally have ActiveState PPM
CF> archives available for Win32 users who want an easy way to install
CF> BioPerl.  I wouldn't have a problem if ActiveState wanted to post
CF> these to their repository.  Why would allowing someone to do the
CF> same for fedora extras be any different?

Hi all,

Given that there seems to be a reasonable consensus (including list
discussion here as well as in private e-mail) from bioperl folks that
including bioperl in Fedora is OK, I'm going ahead and building
bioperl for Fedora >= 6 (it's currently in the development branch).  I
thought about the issue carefully and this seems to makes sense for
several reasons:

1. Biopackages.net isn't currently building packages for Fedora Core 6
   and later (as Allen said, that may happen later when more build
   resources come online).  I won't build perl-bioperl for FC-5 or
   earlier to make sure that the Fedora package doesn't disrupt
   installations with the biopackages.net version.

2. Currently I've only run the the base bioperl (live) package through
   the reviewing gauntlet, but I plan to add the bioperl-run package
   as well.  Even though the bioperl-run package is intended to use
   third party packages (e.g. Clustal etc.) which may not be
   distributed with Fedora, it appears that the bioperl-run package
   contains code that can download those packages directly (albeit
   outside the RPM package system).  And some of the external tools
   could be packaged in Fedora because they have open-source licenses
   (e.g. Wise2, EMBOSS, NCBI toolkit etc.)

   Furthermore it appears the biopackages.net version of that package
   doesn't actually have "Requires:" that would automatically install
   those third-party tool that is run via bioperl (e.g. Clustal) in
   any case, so when biopackages start building for >FC-6 the Fedora
   perl-bioperl* packages can function as a drop-in replacement
   without disturbing other biopackages dependencies such as genome
   databases.

3. Third-party packages that can't be included directly in Fedora
   (such as Clustal) that can be used by bioperl-run could still be
   added via third-party repos like biopackages.net, in the same way
   that the multimedia packages gstreamer and gstreamer-plugins-good
   live in Fedora, but gstreamer-plugins-bad containing patent
   encumbered MP3 codecs with live in Livna.

Cheers,
Alex