On Mon, 2 Sep 2002, John wrote: > I know, people have already said that. However, it yields > results that are clearly wrong. Nobody would think version > 0.001 was the same as version 0.1. ehh??? If I am doing development, using machine assisted numbering, and find versions with: 4.011 and 4.01000 and 4.010 I might conclude: 4.01000 and 4.010 are identical, because they float to the same value and 4.011 is later -- right? But maybe we should split on the radix point, and toss out leading zeroes. so: we see: 11, 1000 and 10 in my example. Now, 4.011 is earlier than 4.01000, for clearly 11 is less than 1000 -- or wait a moment ... ... unsolvable with a single ruleset. > Somewhere there also needs to be something that flags "This package is > compatible with Red Hat Linux xx.yy" so that Mandrake, SuSE and others' > packages will not satisfy dependencies unless rebuilt. > > That may reduce the need for "mdk" and such in package names. ... <smile>. Always in there pitching, John. This may be a matter for rpm-list. It exists already, after a fashion. Perhaps the world would be a better place if instead the vendor tag, specified in /usr/share/doc/rpm-4.1/Makefile, were USED by Red Hat in all package names, similar to an IANA number OID brand: ... RPMCANONOS = linux RPMCANONVENDOR = redhat RPMCONFIGDIR = /usr/lib/rpm ... as found in /usr/share/doc/rpm-4.1/Makefile ... but it may make sense to abbreviate the form to 'rh' Or everyone else in the world could be reasonable, and simply conform their behaviour to the version resolution code and agree to let another decide their versioning. <smile> Cheers, -- Russ Herrold _______________________________________________ Redhat-devel-list mailing list Redhat-devel-list@redhat.com https://listman.redhat.com/mailman/listinfo/redhat-devel-list
