On Sat, Jun 6, 2009 at 6:37 PM, Shawn Walker<swalker at opensolaris.org> wrote: > Peter Tribble wrote: >> >> It's essential that the user be able to force an override of the >> dependency checking, > > No, it is not essential.
Of course it is. Dependencies aren't really anything more than hints - giving them some mystical significance and elevating them to a position of such importance that their word is law just gets you into a rathole. If you had a closed ecosystem, all packages were factored perfectly, all dependencies were perfectly accurate, no packages or software contained bugs, and every single customer/user had exactly the same requirements, and you understood all possible installation and deployment scenarios, then it might be reasonable to argue that relying on dependencies was a sane thing to do. Unfortunately the real world doesn't satisfy any of those constraints. > There are numerous mechanisms already available to users to either correct > or deal with broken packages. So, specifically, how do you uninstall a package that is blocked by a dependency that you know to be inappropriate. > Ultimately, it is not reasonable to expect a package management system to be > able to correctly manage broken packages. So that's why it needs to let us help it. > I believe that --force options are ultimate a cop-out; they fail to deal > with the real, underlying issues. Unfortunately refusing to allow such flexibility doesn't magically make the real underlying issues go away. -- -Peter Tribble http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
