After a few years of experience in using tla I want to make an observation about the arch version namespace that was sparked by the most recent comments about the 3-part names being "dictatorial."
Let me try it this way: the namespace structure, while a good idea, doesn't go far enough to fulfill its potential in the spirit of the "library" analogy that Tom likes to use. Just to belabor the analogy a little: I don't put Dewey decimals on my personal library at home. It's just not an appropriate way to organize my books. I use something simpler. What if the namespace syntax was defined by a per-archive regexp. The default could be the usual arch namespace. But I'd also like to see an additional semantic enforcement relating to common namespace components and version relationships (via "tag", which I also think should be renamed to "link", to reduce severe overloading of the term). Currently, this is entirely unenforced. foo--A--0 and foo--B--0 need not be related in any way, although one might assume that they are since they only differ by a "branch" component in the namespace. It's a little like being able to put an art book in the QA section, and having no librarian to tell you you shouldn't do that. If I have an archive of one-off test programs of a dozen lines or so each (something I do have, actually) that I use to prove little things to myself, this is like my home bookshelf, and I don't need a Dewey system for it. I'm not going to make branches or have major versions of these things. A simple freeform namespace is best here. I want to drop in a predefined namespace syntax and semantics in this archive which is basically "freeform" and "no semantics." I think the tla experiment in "one namespace for all", while a worthy one, has fallen short. We're still talking about whether branch should come before number, or vice-versa, and people are still unhappy with carrying around and dealing with metadata that they don't need. (I don't release software versions. Do I really need to carry around "--0" for rest of my life?) I think there's lots of domain specific reasons for alternate namespaces to be appropriate. Anyway, just some thoughts. Maybe this will resonate with others to start a more serious discussion. Bob _______________________________________________ Gnu-arch-users mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnu-arch-users GNU arch home page: http://savannah.gnu.org/projects/gnu-arch/
