On Tuesday 03 June 2008, Theodore Tso wrote: > On Tue, Jun 03, 2008 at 10:54:14AM +0200, Frans Pop wrote: > > Does that mean that the fact uuid-runtime is _not_ installed by > > default on new Debian systems is deliberate (the second part of the > > BR)? > > Thanks, I didn't notice the second part of the bug report. Hmm. That > can go either way; there are arguments in both directions. The uuidd > daemon as designed is not strictly required; it's just that if you are > going to be generating time-based UUID's at a high rate (which a > certain commercial Enterprise Resource Planning system does when > installing its database) without the uuidd daemon, it is possible to > end up with duplicate UUID's.
Seems to me that can (and should) be solved by making any application that does require the daemon for that reason depend on uuid-runtime. > The flip side of it is, first of all, time-based UUID's are not the > default; random UUID's are [...] Good. Strong argument for not installing it by default. > The other potential reason why we might want to install uuid-runtime > by default is that I moved uuidgen into that package. It makes sense, > as it *is* a runtime program, and bundling it with either libuuid or > e2fsprogs probably was less intuitive --- but over the years, some > people have gotten used to the fact that uuidgen was always present. > So it has been a bit of a surprise for some folks that uuidgen isn't > present any more, and it hasn't been obvious to some folk where to > find it. That means that it is probably correct to leave it as a recommends so that uuidgen does not get "lost" on Etch->Lenny upgrades. But IMO your explanation is support enough for not including it by default on new installations. Given that the package has uuid in its name and that uuidgen is clearly mentioned in the package description, it should not be too hard to find for people who want it. Other packages that use uuidgen should just add a dependency. Are you aware of any that do and that might be missing the dependency because their maintainers are not aware of the split. So, my suggestion would be to leave things as they are. Thanks for the elaborate explanation Ted. This also nicely documents the decision. Cheers, FJP
signature.asc
Description: This is a digitally signed message part.
