[Pkg-scicomp-devel] Package categories

Michael Banck mbanck at debian.org
Sat Aug 2 11:41:54 UTC 2008

On Thu, Jul 31, 2008 at 02:42:28PM +0200, Ondrej Certik wrote:
> On Thu, Jul 31, 2008 at 2:26 PM, Christophe Prud'homme
> <prudhomm at debian.org> wrote:
> >>> At the same time, I am wondering why abinit got uploaded to pkg-scicomp
> >>> without consulting the debichem team first?
> > simple and honest answer:  we didn't know about debichem. the person
> > who brought abinit to pkg-scicomp, Ondrej Certik, does a superb job
> > and he is certainly very helpful in bringing good scientific software
> > to Debian. I suggest that you contact him and ask him whether abinit
> > maintenance
> > should be done in debichem.
> I didn't know about this group. If you want to help with the package,
> we can add debichem to uploaders.
> As Christophe said, I also don't follow all lists, it's just
> impossible. I follow debian-devel from time to time, then debian lists
> and if I want to package something, I do ITP bug and then bring it to
> the team that I find the best. So if you want to help with a package,
> imho you should talk to us, or reply to the ITP bug. And I would say,
> yeas, thanks, go ahead with the packaging.
> So, if you want to help, I have no problems if you start taking care
> of the package from now on. For example there is a new upstream
> release that should be tested and uploaded.

I'm certainly willing to help in general, but I think it makes sense to
discuss the scopes of the different project in more detail.

We have the following, as far as I can tell:

biology -> debian-med
chemistry -> debichem, with some in debian-med and debian-science
physics -> (?, right now pkg-scicomp and debian-science, AIUI)

That is for the applications.  Then there is the underlying
numerical/mathematical support packages like blas, lapack, etc. which go
into pkg-scicomp mostly.

My personal opinion would be to package end-user applications either in
a specialized team/repo like debian-med or debichem, or, if no such team
exists for that field of science (like physics, as I understand it) in
debian-science.  Packages (mostly libraries), which are of general use
to scientific packages would be packaged by pkg-scicomp.

At this point I assume that it is no problem in all of the above
mentioned (alioth) teams that members from one team get added to the
other team once we established that a particular package that person
maintains should get moved over.  

One issue is the DVCS the team uses, some seem to use git and some svn,
so we should see how to collaborate here.

Now, for the more specific case of abinit (and openmx and dft++):  I
looked at their homepages a bit more, and I think it is probably useful
to draw the line between code which does periodic (i.e. mostly solids)
systems to ones which have concrete coordinates and basis sets (i.e.
isolated molecules in the gas phase).  So the former would more physics
oriented, while the latter is more chemistry.  

So I'm in favour of moving abinit and openmx to debian-science (or a
physics-related packaging repository).  Doing the same reasoning, v-sim
should be moved from debian-science to debichem (or debian-med).

All this can wait until lenny is released I guess, but agreeing on a
general way forward now (and including that plan in the "Bits from
debian science" newsletter) can be done now.



More information about the Pkg-scicomp-devel mailing list