#234: track installed files to allow uninstall ----------------------------+----------------------------------------------- Reporter: duncan | Owner: Type: enhancement | Status: new Priority: normal | Milestone: Cabal-1.6 Component: Cabal library | Version: 1.2.3.0 Severity: normal | Resolution: Keywords: | Difficulty: hard (< 1 day) Ghcversion: 6.8.2 | Platform: ----------------------------+----------------------------------------------- Changes (by claus): * cc: claus (added) Comment: (following Duncan's suggestion to record my opinion on the importance of this ticket) When I suggested "uninstall", 2 years ago (#106), I was very surprised that it wasn't rated as both urgent and important. The *only* reason for delaying it is that it is not something you'd want to get wrong (mostly: removing stuff you're not supposed to touch; though leaving garbage all over the place isn't nice, either). The very first thing I check about new software is how to get out of it after I start it, and, having had some trouble with nearly-impossible-to- uninstall software, I've come to see software that installs itself without providing for proper uninstallation as deeply unprofessional, the never- use-their-stuff-again kind of experience (installing everything into a single new directory that can be removed without causing damage is okay as a minimum requirement, but if you do anything else, you'd better know how to undo it). Doing temporary test-installs before releases (to ensure that packages are complete), forgetting/changing installation prefixes (Cabal's handling of which differs between platforms), having different compiler versions with different Cabals, Cabal evolving and putting things in different places, more and more packages on Hackage that people will want to install to try them out - all reasons why uninstall support is needed before spreading the news about Hackage/Cabal leaves potential Haskellers with a bad taste (not to mention that left-overs from previous installs bring the risk of inconsistent installations). I really thought the importance and urgency of this ticket was self- evident but, if not, I hope this explains my point of view?-) Btw, there seems to have been some relevant discussion in this thread http://www.haskell.org/pipermail/cabal-devel/2008-May/002907.html concluding, I think, that some refactoring would be needed before tracking could work. -- Ticket URL: http://hackage.haskell.org/trac/hackage/ticket/234#comment:3 Hackage http://haskell.org/cabal/ Hackage: Cabal and related projects