James Carlson wrote: > It should not be the case that there are special things that can be > done when the system is installed, and that cannot be done at other > times. In fact, the process of setting up these bits of configuration > shouldn't be different. (E.g., setting up name services.)
sysidkrb5 and kclient being my personal bug about these being different. I pushed hard for kclient functionality to be done in sysidkrb5 but I got out consensused (yeah I know that isn't a word but there wasn't a vote this was ARC fast-tracks). The other thing about sysid is that the bits need to be able to be configured and unconfigured independently of each other and unless they really really require a reboot they shouldn't do it. So I'm 100% with James here, configuration can be done as part of install but it must be 100% functional after install too in exactly the same way. In my definition (and I assume the one James uses) Configuration does not imply picking which packages to install, only configuring things in those packages that have been selected for install. This kind of implies that we should run sysid AFTER install, or at least after package selection, not before it like we currently do. One other data point is that I believe this is the model that most if not all of the products in the JES stack are either doing now or are moving towards. -- Darren J Moffat
