We have a tool here at Edinburgh University (Scotland) that we use to keep our machines up to date with rpms. It's operating model is quite unlike any of the package managers out there I've seen so far. It's not as advanced in terms of dependency resolution or remote repositories, but - it has one very very useful side effect because of the way it works: you specify a file per machine such as: <example file> foo-12.5.8-4 thingy-1.2.3-* other-thing-*-* </example file> (* are wildcards that say install latest version found) .. on a central configuration server. The tool then runs at specified intervals and makes the machine match it's package profile. The config files have a CPP like #include system too so common rpm sets can be shared between machines. Now this is the interesting bit... during term we can have rpms in the lists like teaching-software-1.2.3-* and then on the first day of the holidays remove it all and add quake by simply removing one package from the list and adding another. .. the lab changes from a work room to a play room all by itself in a matter of minutes. Several research projects we have been invoved in have applied this model to be able to on demand reconfigure grid systems. Now that only does software and not configuration, but we have a tool for that too but that's another story. What I'm wondering is if anyone else would find a model like this useful? From our experience I'd say that our system is lacking in some areas - we at the moment have to specify _all_ the rpms needed. Yum's rpm-group capabilities and automatic demendency resolution though have lead some of our team to start thinking about pooling out efforts into a more generalized rpm updating tool. I guess this is just a feeler to see if the yum community would be interested in looking at this a bit more? Carwyn --
