Version control systems
dons at galois.com
Sat Aug 9 15:38:41 EDT 2008
> > Ian Lynagh:
> > >On Fri, Aug 08, 2008 at 12:04:15PM +1000, Manuel M T Chakravarty
> > >wrote:
> > >>
> > >>I seriously hope the plan is to move all *core* libraries (including
> > >>GHC's cabal repo) etc over to git, too. In other word, everything
> > >>that you need to build the development version of GHC should come via
> > >>git. Having a mix of VCSs would be the worst option of all.
> > >
> > >No, the plan is to move only the GHC and testsuite repos to git, as
> > >the
> > >others are also used by hugs, nhc98, etc.
> > >
> > >It would be possible to move GHC's Cabal repo over too, as that is
> > >private to GHC, but given the other libraries will be using darcs
> > >anyway
> > >I think it is simpler to keep all darcs repos using the same VCS.
> > I think all *core* libraries must switch. Seriously, requiring GHC
> > developer to use a mix of two vcs during development is a Very Bad
> > Idea.
> I agree with this.
> As Audrey says, you have to lower the barrier to entry. That means:
> * one build system
> * one vcs
> to build ghc (and anything it requires, such as the core libraries).
> This is a chance to make a big step towards accessibility, let's make that step.
I just want to add to this. We're offering an unusual product: a lazy,
purely functional language. This already separates us from the
mainstream. So how to we ensure we minimise the stress of adopting
something so unfamiliar?
By ensuring that in all other respects the environment they have to
learn is familiar and simple.
I think we risk isolating oursevles yet further -- and creating new
barriers to adoption, beyond those we can't avoid -- by adding more
complicated dependencies (two revision control systems, one of which,
darcs, is now firmly out of the maintstream).
Instead, if we just use ubiquitous, common tools -- like git -- for
everything, we minimise the pain for people, and sit firmly in the
mainstream of open source.
If anything has been learnt by Spencer and I while working on xmonad,
dependencies, dependencies, dependencies
reduce these, and you gain eyeballs, and ultimately developers. Increase
these, and you end up isolated and marginalised.
So let's capitalise on this switch to git, and take the opportunity to
remove one big dependency from the system.
More information about the Glasgow-haskell-users