Hi Trygve, Am 01.02.2009 um 21:33 schrieb Trygve Laugstøl: >> Well, we have multiple namespaces: >> >> (1) directoryname inside GAR >> (2) catalogname of package >> (3) packagename CSW<name> >> >> The following statements should be valid: >> (a) For most packages the should be (2) == (3) with some exceptions >> (e. g. Perl modules) > > The Perl modules getting prefixed with "pm_" is fine, but there are > other discrepancies which I really don't like. But I guess that's > life.
If there are really ugly ones please make notes of it so we may address it some day. >> (b) By design we have 1:n for (1):(2) and (1):(3), because one GAR >> description may produce >> several packages > > True, but most builds create a single package (not counting _doc etc), > so it should be 1:1 in most cases. Yes. And if it is not 1:1, (1) should be a prefix of (2) >> (c) There should be at least one package with (1) == (2) >> >> For you there is >> >> (1) collections >> (2) commons_collect and commons_collect_doc >> (3) CSWajccollect and CSWajccollectdoc >> >> and hence neither (a) nor (b) nor (c) holds. >> >> I propose to use >> >> (1) GAR directory = ajccollect >> (2) catalogname ajccollect and ajccollect_doc >> (3) CSWajccollect and CSWajccollectdoc >> >> Having a hard-to-remember name is imho worse than having pkgname != >> catalogname. > > Yes, I have to agree with that. I just hope that it can get a better > name that "ajccollect". How long can a name be? IIRC there was a limit on the package name of 11 characters ages ago which was then raised to 256. There is a cosmetical limit on the catalog name, so it can be pretty printed on e. g. 'pkg-get -a' > "commons-collections" is > also one of the longer names, most are somewhat shorter. Yes, that is too long. >> Phil, how about adding this to the standards? > > +1 Am 01.02.2009 um 21:34 schrieb Philip Brown: > On Sun, Feb 01, 2009 at 09:25:25PM +0100, Dagobert Michelsen wrote: >> Phil, how about adding this to the standards? > > not sure which "this" you mean. that was a long email :-) Well, the mapping stuff about package- and catalog-names and that they should match and that they should be prefixed with -doc etc. Like what we discussed some time ago: > Am 30.08.2007 um 18:41 schrieb Dagobert Michelsen: >> Hi Phil, >> >> Am 30.08.2007 um 18:28 schrieb Philip Brown: >>> I agree splitting off "rt whenever it makes sense". >>> I disagree that it makes sense, for libart ;-) >>> >>> "_rt" is normally used, when there is a separate, distinct >>> product, that >>> people expect to use. >>> >>> Then, it is sensible to split into >>> product, product_rt, product_devel. >>> >>> The USES for these, would be >>> >>> "product": used for "pkg-get install product" to have something >>> that >>> users actually expect to use directly >>> "product_rt": as a minimal dependancy for other packages that have >>> need of the shared libs, but not the end-user stuff >>> "product_devel": because the devel-related stuff is noticably >>> large, >>> so only people actually compiling/etc stuff with >>> 'product' need to install this >>> Note that in the case of a library, and libart specifically, >>> there is no separate end-user "product". there is nothin that it >>> provides >>> in bin for users to use. >> >> That makes sense. Do you mind adding this to the standards page? That should IMHO go to <http://opencsw.org/standards/build#versioning> Best regards -- Dago _______________________________________________ devel mailing list [email protected] https://lists.opencsw.org/mailman/listinfo/devel
