[Haskell-cafe] Re: 1,000 packages, so let's build a few!

[Duncan Coutts]

On Mon, 2009-02-02 at 10:07 +0000, Neil Mitchell wrote:

> * Some of it comes down to technical issues - for example not having
> cabal.exe bundled with GHC 6.10.1 on Windows was a massive mistake
> (although I've heard everyone argue against me, I've not yet heard a
> Windows person argue against me).

There are reasonable arguments both ways. Listening to some people it
would seem that cabal.exe is there to taunt people with the potential
for usable packaging and then to punish them with broken packages,
quirky dependency resolution and twisty inconsistent installed package
graphs.

There is a tension between getting it out there because it's useful, and
not putting out something that is going to frustrate and annoy new
users. With the current release that balance seems to be pretty fine. I
hope that the next release will make it look a little more mature.

For example look at how many people have been getting themselves into a
complete mess recently from using cabal upgrade. There is a whole new
world of exciting mess you can make when using easy automated tools with
no safety catches. These are not problems that emerged at first because
it was only developers who knew what they were doing who were using it
at first. Now that more and more people are using it we're discovering
these usability problems. I fear that if we'd gone any quicker that we'd
have been completely swamped with bug reports and users might write the
whole thing off. Given the available developer resources I don't see how
we could have done significantly better (except perhaps by releasing
even later).

Duncan