31 Aug
2004
31 Aug
'04
6:40 p.m.
On Tue, 2004-08-31 at 10:00, Chung-chieh Shan wrote:
The rationale for disallowing matching partially-applied type synonyms is that higher-order unification is undecidable.
Higher-order unification is worse than just undecidable (after all, GHC's extended Haskell already includes constructs which are undecidable, which means that sometimes the compiler will loop forever); it's ambiguous. There can be multiple unifiers, none of which is the most general. See my earlier Haskell-cafe message for the trouble this can cause (search for "<technical note>"): http://www.haskell.org/pipermail/haskell-cafe/2004-March/005965.html Carl Witty