On Tue, 2008-10-14 at 11:25 +0100, Magnus Therning wrote:
It seems to me that it adds confusion (two ways of installing things rather than one) while reducing flexibility and 'freshness' of installation.
To me it adds simplicity. I might be developing an in-house tool at work, having all the libraries available in distros makes my work easier.
Yes, we want distros providing as many good Haskell libs and tools as possible. Having Cabal -> native translation tools and doing central package QA seems to me to be the right strategy.
Playing the devil's advocate I'd say that cabal (not the library Cabal, but the tool cabal in cabal-install) is only needed on systems with pacakge managers that are broken or completely missing (e.g. Windows). As such cabal is a waste of time and shouldn't have been written at all; on many systems it's of no use, on the ones where it is useful it's a fix at the wrong level. Somewhat harsh, and not completely in line with my own opinion, but it can be argued that way.
I think they're actually complementary. Sure on Windows it's needed in place of a native packaging system, but even on systems like debian it's still needed in places. It's needed for packages that are too new or are not sufficiently mature or popular to have been packaged yet for the distro. There will always be such packages. Of course many ordinary users would be able to make do with the subset of packages that are provided by the distro and that's great. Duncan