tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: science category

On 04/18/17 09:24, Dr. Thomas Orgis wrote:
Am Tue, 18 Apr 2017 08:54:16 -0500
schrieb Jason Bacon <>:

I'm not sure what your concern is about messiness.  We're talking about
a new category for mostly new packages.  Yes, some existing packages may
be moved to science and dependents will need to be updated, but that's
easy and will probably happen very slowly anyway.
A new category is fine … although my scripts also do not care much and
I have them do a search anyway. One could debate if topical categories
are really that helpful (a scheme as suggested by Kamil would enable me
to guess a package location rather reliably … if we dropped uppercase
letters from names). I hope any moving of existing packages is done
rarely. This breaks local patches, for example … creates work just for
work's sake.

FYI, pkgsrc has a huge potential impact on scientific computing and
research in general.  Most HPC clusters run CentOS, which deliberately

I'm not sure if most clusters really run CentOS specifically, but I am
part of that trend (minimal software costs, maximum hardware delivered,
still somewhat supported by vendors → CentOS). I introduced pkgsrc for
the reasons you mentioned, combining it with differing toolchains
(local patch pending for using external MPI) to present a menu of
userspaces to the users. We want updated software, but also the
unchanged software we used 3 years ago to reproduce computations.
Versioned Pkgsrc prefixes give us that.

HPC is always a niche in terms of user count, but the traces in pkgsrc
are obvious (the whole geography category, for example … and even
things like wip/trinity).

Alrighty then,


There's no rush to move existing packages, so I don't see this as a problem. We can deal with these if and when time permits. I'm more concerned about the impact of "hiding" many new scientific packages under inappropriate categories.

And I think the ability to browse by category is important to many users.

I did a poll among XSEDE campus champions about a year ago and at that time about 80% of HPC clusters were running CentOS or a derivative such as Scientific Linux. Most of the remaining clusters were running RHEL.

I cited HPC and CentOS because pkgsrc serves an unmet need there. However, many scientists who don't need HPC or commercial software (and are probably running Ubuntu now) will also warm up to pkgsrc as the scientific software collection grows. It will simply make sense for them to use the same package manager as their collaborators who have to use RHEL derivatives.



Earth is a beta site.

Home | Main Index | Thread Index | Old Index