installing the package installs the libvirtd service and sets it to start automatically but doesn't actually start it.
Automatically starting services from a %post script is a pure evil & forbidden. The RPM installation could be being done in a chroot for the purposes of building other dependant packages & you defintely do not want this starting daemons.
okay, but would you agree that this is somewhat confusing/frustrating for new users? surely the "chroot for building stuff" is the more uncommon scenario which should be tested for in the script, rather than throwing out the idea because it's evil. would it be okay if the script checked for chroot and started the service if it wasn't chrooted? or are there a million other gotchas queued up behind that one? sorry for pestering/reopening - i'm not after a full explanation (i'm sure that's on the web somewhere) - just a little bit more info. tell me there's a million fiddly little problems that cause the overall "evil" status and that'll be fine - i'll go find some other way of dealing with it. maybe this should be a plugin for yum/pirut/packagekit rather than troubling rpm or individual package maintainers with all the security implications? i'm just thinking it'd be nice if all apps helped the user as much as possible following an install and if we start down that road then the code and rules required by yum would get insanely complicated. better to set a precedent and create a supporting structure to help apps install themselves. why pick on libvirt? well, you'd have to start somewhere and that's where i was when the thought occurred.
Take this discussion to fedora-devel-list if you want to get into it. This is not appropriate for bugzilla.