[Bio-Linux] Bio-Linux future directions - OS choice.

[Dan Swan]

Chris Beck wrote:
> Speaking as a non-grant user I have only one caveat.  It might be wise 
> to also take into consideration the distribution that the bio packages 
> are built-developed under ... if they track a fast moving distro like 
> fedora (which will be moving to kernel 2.6 within the next few months 
> then a slow moving distro like debian might fall behind.  That, of 
> course, is contingent on the pakcages in question being closely tied to 
> the underlying system, if they are loosely coupled then no worries.

In my experience bioinformatics software tends not to track the latest 
jumps in gcc versions too quickly, although this is rarely a problem as 
a quick install of the libstdc++ backwards compatibilty libraries tends 
to fix and issues.  I think there is a general reaction to wait for 
platforms to be seen to be stable, I know I have not made the switch to 
a 2.6 kernel yet.

The Debian install that Knoppix does is actually very up to date as it 
tracks certain elements from testing/unstable (for those who are worried 
about those designations Debian does not describe anything as "stable" 
until it has been around for a *very* long time).

Here are the specs from my test Debian/Knoppix machine:

gcc version 3.3.3 (Debian)
This is perl, v5.8.3
Python 2.3.3
java version "1.4.2_02"
ruby 1.8.1 (2004-02-03) [i386-linux]
Linux knoppix 2.4.24-xfs
Apache/1.3.29

And from my test Fedora Core 1 machine:

gcc version 3.3.2 20031022 (Red Hat Linux 3.3.2-1)
This is perl, v5.8.3
Python 2.2.3
java version "1.4.2_03"
ruby 1.8.0 (2003-08-04) [i386-linux-gnu]
Linux fedora 2.4.22-1.2174.nptl
Apache/2.0.48

As you can see they're very close at the moment.  I think it's extremely 
important to track Perl, Ruby and Python very closely with maintenance 
releases and in my experience Knoppix Debian is much more up to speed 
with this than Fedora.

With regards to the bioinformatics packages, we fully intend to create 
deb packages on i686 machines here, and then have an apt repository (we 
would have done the same for rpm/yum if we had chosen Fedora).  This 
means we will test and version check the software prior to release.
Also things like BioPerl are already packaged as deb's - this removes 
the need for trying to bash through broken CPAN installs and rpm hell 
with GD under Fedora.  Keeping BioPerl up to date on versions 1-3 has 
been a particular bugbear.

I hope this answers some of your concerns.

regards,

Dan

-- 
Dr Dan Swan - Bio-Linux Developer | RHCE
EGTDC, CEH, Mansfield Road, Oxford, OX1 3SR
Tel: 01865 281 658 Fax: 01865 281 696
http://envgen.nox.ac.uk/ | dswan at ceh.ac.uk