Dave Miner wrote: > Karen Tung wrote: > >> I am working on: >> >> 3398 slim_install should be removed from the CD during finalization. >> >> The change is a simple "pkg uninstall slim_install" after all >> the "pkg install" of the pkg list is done. What I want some comments >> on is how to do this more elegantly so that it will be easier to explain to >> the user why things are done this way. >> >> Solution 1: >> -------------- >> Add a section to the manifest so people can specify the list of packages >> to remove >> after install? Perhaps call it <post_install_remove_packages>? >> The DC main app will interpret this and remove the listed packages after >> all packages are installed. >> The advantage of this is that people can easily specify what packages they >> want to remove here. The disadvantages is that it will confuse people. >> They might wonder why we install and uninstall slim_install back-to-back? >> >> Solution 2: >> ------------- >> Do this in a finalizer script. Perhaps add it to the existing >> /usr/share/distro_const/pre_bootroot_pkg_image_mod. >> That script currently only removes the IPS indices. Since slim_install >> is also another packaging artifact, we can specify it as an argument >> to the finalizer script, and inside the script just call "pkg uninstall >> slim_install". >> For the globalization CD, babel_install will be specified as the argument to >> that script. >> >> Personally, I like solution 2 better, because I don't think removing >> packages >> is a common thing to do in the future. We just have to do it in this case >> because we want to clean up some packaging artifact. >> >> > > Those creating custom distros, which are likely to be minimized in a > variety of ways, can take two approaches: > > - start with a core that's less than you need, and add to it > - start with something larger than you need, and subtract from it > > Or some combination of the two, if you have multiple group packages that > represent the desired function. Something like slim_install minus > firewire if you have no hardware that supports it, plus ss-dev minus > gmake if your standards are to use Sun make. > > Each is valid. The scenario with slim_install and babel_install is but > one specific instance where we do this for a specific reason. > > I would much prefer something along the lines of #1. > > Dave >
Thanks for your input Dave. What you said makes a lot of sense. I will be implementing option #1. --Karen
