Alok Aggarwal wrote: > > On Fri, 20 Feb 2009, Dave Miner wrote: > >> Ethan Quach wrote: >>> >>> >>> Jens Deppe wrote: >>>> Hi >>>> >>>> Should I expect that a manifest without any packages defined, when >>>> used in an install, would result in the base set of packages being >>>> installed (SUNWcsd, SUNWcs, slime_install and entire)? >>> >>> No, that list isn't baked in anywhere, so the manifest needs it there >>> for now. I believe there's an RFE for something similar to what >>> you're asking about, however we really need to define what >>> constitutes a "base set" of packages for installation; and perhaps >>> different sets for different types of installations - applicance, >>> minimal install, etc.. >>> >> >> I'm doubtful that installing some hard-coded package list when none is >> provided is at all a good idea in terms of evolving the technology, or >> in using it with multiple different distributions. Defining standard >> package set names is useful, and providing examples of using them as >> well, but I would prefer to regard a manifest which didn't specify any >> packages as incorrect. > > So, I understand that having a hard coded list of packages > that gets installed by default (as in, when no packages are > specified in the manifest) is not going to be ideal across > different distributions. > > But does that point to the need for having a different set > of defaults on a per-distribution basis or not having defaults > altogether? > > In some ways having a default for the package list > is similar to having a default target disk (which we obviously > do pick). So, why get rid of default packages altogether? >
Perhaps it's a matter of "degrees of defaultedness". Shipping a sample manifest which has a default package list that's identical to the live CD is OK, and what I asked for originally. Having other manifests which do not specify a package list somehow default to using that same package list seems over the boundary, in my view. Dave > Alok
