On Tuesday 04 September 2007 14:43:41 Andreas Tille wrote: > On Tue, 4 Sep 2007, Steffen Moeller wrote: > > * When thinking about automated installations of software (i.e. in grid > > computing) we need a language that allows us to talk about what is > > eligible for installations and what is not. Debtags are not perfect and > > other efforts describing various kinds of properties that software can > > have, there is nothing as sweet as Debtags to talk about what the > > software is actually doing. > > Well, I'm convinced that DebTags might be a very great tools for different > things, but I doubt that it is the best idea to base installations of > clusters on DebTags technology. You certainly want to know _exactly_ > what is installed on your cluster and do not really want it to be changed > by any change in the DebTags database. I think for this purpose the > meta package approach is the better way to go. The debtags would be used as constraints: "I would allow the requested installation of program X if there is a facet f such that f(X) is stated in debtags and f is among a set of facets that characterise my cluster because I say so. You would _not_ use debtags to install every X with f(X) in debtags for some f in F. > > * Debian integrates communities. This is my way to read Custom Debian > > Distributions that are basically saying they people flock together to > > extend Debian towards a particular direction. Specialisation of Debian > > comes with a specialisation of terms. It is natural. > > Sure. But I think subsetting makes sense in case your main set is > to large to be managed with the means you have at hand. IMHO this > is actually not (yet) the case. We want to extend Debian but I don't > think that we should try to make a science out of classifying and > subsetting what finally might end up on a real live installation > all together again. Hm. But you saw from Benjamin's reply that he found it not natural to have 3D structures a separate facet, and I do not think we should spend much time on such easy decisions: of course we need that, but how could Benjamin know? I am in favour of some decentral managing. The technology may be there to have a shared maintainance but it would be less efficient and probably consequently also less fun. We could have something like: categories with more than 5 entries get into the main debtags distribution ... or something alike. > > I like the above sketched suggestion to allow for disjunct sets of facets > > that are maintained by different communities. It would seem natural to me > > to eventually allow for sub-facets of some kind with a higher number of > > ":" in their IDs to thus allow for an easier reduction of complexity. > > Though ... well ... it may not be needed tomorrow. > > I think we could wait with our fine grained subsets until this is > implemented. Once this is done also the number of packages that > rectifies a more fine grained subsetting will have increased. :) Fine. Steffen
Attachment:
signature.asc
Description: This is a digitally signed message part.