Re: debian-science repository structure
>>>>> "Sylvestre" == Sylvestre Ledru <sylvestre.ledru@inria.fr> writes:
> Martin Prinz wrote:
>> OK. I just thought this was due to some misunderstanding by the
>> people who checked those in. My fault! Thanks for your work an
>> that!
Sylvestre> No worries, David Bremner has to decided which category
Sylvestre> would be the best.
[sorry that I am replying to the whole thread at once, attribution
could be better here]
At the moment I am preoccupied with deciding how to best transition
my quilt based packages to git. Of course keeping using quilt is an
option. As someone who is not a debian developer, and needs to have
his packages sponsored, I am concerned with not making the sponsorship
process any more difficult. Right now I am playing stgit, with the
idea of exporting a stack of patches into debian/patches at
source-package build time (i.e. using something like git-buildpackage
or gitpkg). I'd like to hear about any related experience that people
have.
On the topic of categories, I also prefer not trying to force the
packages into a hierarchy. I think every brain cycle we spend on
deciding whether a package is physics or chemistry could probably be
used for something else. This is not to take away from categorization
efforts, just to claim that the fit with a strict tree is not exact.
I think that tasks sound like a reasonable primary format for the
metadata. One can generate static web pages (I guess debian-med
already has tools to do this). I think it should be possible to
generate mr configurations, or git superprojects automagically from
the tasks files as well. I guess we need to include appropriate
VCS-git and VCS-browser fields in the tasks files. I would be curious
to know what a use case for "check out all of the physics packages"
would be.
All the best,
David
Reply to: