On Wed, 22 Sep 2004, Garrick Staples wrote: > On Wed, Sep 22, 2004 at 05:12:38PM -0400, Robert G. Brown alleged: > > I don't know why it has taken me so long, but it finally occurred to me > > that it would be just smashin' brilliant to use yum as a distribution > > mechanism for some GPL programs I have written and maintain and that are > > used by various groups. After all, any website (including a personal > > one) is a potential yum repository, and it can take the guesswork out of > > the entire update process for remote users. They can either mirror the > > mini-repository or point yum at it on their systems. > > The biggest problem here is that distributing _binary_ packages from upstream > doesn't work very well. Each distro has too many differences that prevent such > a distribution of a complex package. The upstream source would need to > maintain their own build systems for every supported distro. > > Btw, Macromedia's flash module is distributed in such a way. > http://macromedia.mplug.org/ I believe that mozilla has such a distribution mechanism set up as well. I don't >>think<< that this particular (relatively simple) application set would be terribly conflicted (and the source rpm's would be there if the binary rpm's failed to install). It requires libc, libxml2, ncurses and I think that's about all. libc is likely the least stable of the three -- ncurses is more or less static in features and API, libxml2 changes a bit inside the API but I haven't encountered a killer bug post libxml. > <idlethoughts> > Imagine a set of publicly available build hosts for all distros and > archs... now imagine Debian :) > > For open source software, it might be fun to have yum build src.rpms directly > from upstream. Oh wait, that system is already called Gentoo... anyone want > to port portage to RPM? > > Coming full circle, the whole idea of a "linux distribution" is to have a > _single_ set of matched packages. > </idlethoughts> Yeah, but for a developer/maintainer of a small package or set of packages, the best one will ever do is distribute source rpm and perhaps tarball on the side for "source", and a binary rpm built on one or maybe even two whole distributions (whatever the developer has access to in a stable form). Testing has likely only been done on those distros as well. FC seems like a horse one can ride -- the whole thing is fairly "dynamic" and close enough to the edge that riding it one won't fall critically behind. Of course that means periodic growing pains. I'm still getting FC2 (really 2.6.x kernel) derived buglets out of my packages. This is going to be more or less an experiment. I think the package is already in gentoo, but I don't do the build testing there so I'm not certain if it works or stays current. I'll let y'all know how it turns out. rgb Robert G. Brown http://www.phy.duke.edu/~rgb/ Duke University Dept. of Physics, Box 90305 Durham, N.C. 27708-0305 Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb@xxxxxxxxxxxx
