I think all this discussion about numerics in Haskell is great. I'm convinced
that designing good libraries is a major creative act, not just an add-on to a
language; and that the existence of good libraries has a big effect on how much
use a language gets. ('Good' means both having a well-designed signature, and
at least one efficient implementation. Sometimes these two conflict.) Library
design is sadly under-valued: people who design and implement good libraries
should get at least as much credit as people who design good languages.
The trouble is that it's easier to identify flaws in current libraries
than to figure out better ones. Certainly in the numeric area there sems
to be a big tension between completeness and simplicity.
Bottom line: Haskell 2 is pretty wide open. If some group of you
out there get excited about this and design and implement a coherent
set of numeric libraries then I think you'd be welcomed with open
arms, even if it wasn't compatible with Haskell's current numeric
class structure. Sergey's work is moving in that direction, which
is great.
Above all don't leave it to us compiler hackers! We aren't
numeric experts like some of you are, and are most unlikely to do a
good job of revising Haskell's numeric libraries.
Simon
