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

Re: debian-science repository structure



Am Freitag, 16. Mai 2008 15:57:59 schrieb Frederic Lehobey:
>   (As a maintainer of this tasks package) I would like to emphasize
> that the names were chosen from existing debtag, namely
> [...]
>   The goal is the have the simplest common (even if not perfect)
> namespace. I have renamed the corresponding wiki pages
> (http://wiki.debian.org/DebianScience) accordingly.
> 
>   I advocate, when creating new categories in the repository, to take
> a name already in a field::* or use::* (or other) facet of existing
> debtags).

Frederic, thanks a lot for your input on that!

Looking at the current situation, the repositories are quite unstructured: The 
SVN repo contains the tasks created by Sylvestre and some packages checked in 
at the same level which is quite confusing. Git repo does not contain the 
tasks, just packages at the first level. I was thinking about that situation 
today and would like to share my view on that with you, looking for your 
input:

Since we do not know if we will move to one repository or keep both running 
forever, I'd prefer to have the same repository layout in both VCSs.

Moreover, tasks seem to be something "artificial" and problematic to me. I see 
the need to group packages into tasks, especially for the meta-packages. But 
while working at a package at the moment I was not able to decide where to 
put it: physics or chemisty? It would fit into both tasks perfectly. So I 
thought about why we should add tasks to our directory structure and did not 
see the point. It's an artificial grouping which can be done by debtags or by 
the really great solutions with task files, as in the Science CDD 
directory[1]. I do find a structure like

-- debian science
   +-- homepage
   +-- packages
        +-- bar
        +-- foo
   +-- policy
   +-- tasks
   +-- ...

a lot more convenient. I think of packages to be just some aspect of Debian 
Science and therefore would like to split it from the rest. There should be 
no deeper structure since the packages can be grouped by the task files. It's 
trivial to get a list of all packages belonging to a task, and it can be made 
even more trivial with writing a tool. (I just doubt that working on all 
packages in a task at the same time is a common case.) The other directories 
contain things that the group developed over time, such as a policy[2], the 
task files, homepage, scripts and what may come in the future. I also see the 
problem of moving a package from one task to another or fitting two or more 
tasks, and I see no solution to that despite handling tasks seperately from 
packaging.

I hope you find my concerns valid and I'd be happy if you'd all share your 
opinions on that with me.

Best regards
Manuel

[1] 
http://svn.debian.org/wsvn/cdd/projects/science/trunk/debian-science/tasks/?rev=0&sc=0
[2] I already started to work on that but it's not uploaded yet. I'll announce 
it as soon as it's something readable.


Reply to: