Non-H98 crusade, contd.

[ross at soi.city.ac.uk]

On Sat, Feb 26, 2005 at 08:19:06PM +0100, Sven Panne wrote:
> From the discussions in recent years it is far from clear to
> me what would be an accepted extension. The current type system
> extensions interfere with each other and it looks like some can be
> emulated by others. Furthermore, nhc98 implements almost none of these
> extensions, but I guess Malcolm would happily accept any volunteers. :-)
> Furthermore, there's still hbc, tools which operate on Haskell sources,
> etc. etc. Keeping the non-H98 part of the standard libraries low is
> still a worthy goal, even in 2005...

> Note that I'm not against any extensions, I'd be more than happy if a
> new revision of H98 would emerge, but I seriously doubt that simply
> declaring what 2 implementations provide as a standard is the right
> way to proceed. Perhaps we should make a few more carefully selected
> addenda to the H98 report.

Indeed.  Of all the extensions implemented by both GHC and Hugs, the only
ones that seem ready are

- rank 2 type signatures, and

- polymorphic components for data constructors (giving them rank 2 types).

The extensions to the type system are well understood, the operational
semantics is unaffected, and they get you useful things like runST,
Church encodings and polymorphic continuation monads (like ReadP).
Though polymorphic components do complicate the denotational semantics.