Am Montag, den 19.05.2008, 00:07 +0200 schrieb Andreas Tille: > On Sun, 18 May 2008, Sylvestre Ledru wrote: > >> Those are of course valid points, though I still think that is something that > >> can be generated from the meta-information that is already available. Writing > >> a script for that should be fairly trivial and I'd be happy to do so! > > Does anyone else has an opinion on this ? Andreas ? Charles ? > I admit I do not really understand the question. What meta-information are you > actually talking about. I've thought we would talk about categorisation of a > package and I don't think that it is in any of the meta information of a package > (except perhaps in the Vcs-* fields *if* we just decided for a category and thus > the problem is recursive). Sorry for being unclear! As stated in my other email, I was talking about the task files. Those are (in my understanding) meta-information about which packages belong the task or category and therefore which tasks or categories a package belongs. For example: "tasks/physics" mentions "foo" and "bar", "tasks/chemistry" mentions "bar" and "baz", then "bar" is in the physics and chemistry task. There is no need to put a directory structure to the repo for that. I really like the idea of the tasks approach that is used in Debian-Med / Science CDD! That's also the point I was trying to make: I do not object tasks, I just object a task-based directory hierarchy. If one wants to download all physics packages, this is trivial: Parse that tasks/physics file for the package names and pass each package name to debcheckout. Done. Writing a tool for that is a piece of cake and this is aproach is IMHO superior than to checkout a part of the VCS tree that might not contain a package because someone put it into a different category. Sorry for causing confusion on that! Best regards Manuel
Attachment:
signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil