Ross Paterson:
On Mon, Aug 11, 2008 at 04:17:59PM +0100, Simon Marlow wrote:
The main obstacle with just switching the core libraries is that they are shared by other implementations and other maintainers. So I see no alternative but to create forks of those repositories for use by GHC, unless/until the other projects/maintainers want to migrate to git.
Forking is much worse than using multiple vcs's, and if we don't fork, anyone working on those libraries will have to use git at least to get GHC HEAD to check that they're not breaking it. And clearly GHC developers outnumber developers of other implementations. (I don't think a move to git will lead to more GHC developers, but I buy the interns argument.)
Ah, good point! Changing ghc to git means *all* developers of boot libraries need to use git *regardless* of what repo format the boot libraries are in. After all, they need to validate against the current ghc head before pushing. In other words, the decision to move the ghc repo affects all core library developers anyway. No use pretenting that changing only the ghc repo (and leaving the rest in darcs) would make anything simpler for anybody.
My concern is that there are rather more developers of libraries and assorted other packages, and this will place an arbitrary divide across those. Unless everyone moves to git, of course.
There are surely more developers of libraries in general than there are GHC developers. However, I doubt that there are more developers of boot libraries, who are not also ghc developers, than there are ghc developers. The change doesn't have to affect anybody, but ghc developers and *core* library developers. Manuel