On Tue, 30 Dec 2008 16:30:28 +0200 George Danchev <danchev@spnet.net> wrote: > > "some sort of API version discriminator" doesn't sound as if you've > > understood SONAME transitions. > > ... or you better understand [1] that you should avoid keeping SONAME > artifacts in the -dev package names, thus avoid changing -dev package name on > each SONAME bump, which would make release team cry upon transitions, loudly. Which is why, generally, I prefer to use libfoo-dev - it isn't an argument (to me) for using some number other than the SONAME in the -dev package name. It would be particularly confusing to use a number in the -dev package name that is just "some sort of API discriminator" but that had no relation to the actual SONAME. If a number is used, it should change in step with the transitions and be predictable from objdump -p. However, once a transition does come along, if you want to retain libfoo1.2-dev and libfoo2.0-dev, then it makes more work for some but allows libraries with hundreds and hundreds of reverse-dependencies to have a sensible migration path. Such libraries are few and far between, thankfully, but the ability to retain the SONAME in the -dev package name (and the source package name) is important for a small number of fundamental libraries. There is then the inevitable pain of deciding that libfoo1.2 simply has to go away at some point. :-) Where would we be without libc6-dev ? -- Neil Williams ============= http://www.data-freedom.org/ http://www.nosoftwarepatents.com/ http://www.linux.codehelp.co.uk/
Attachment:
pgpb2pfPiCDMh.pgp
Description: PGP signature