(via The Types Forum)
Martin Odersky:Alan Jeffrey has shown in a previous note to the types mailing list
that the original scheme for local type inference used in GJ was
unsound. The unsoundness was caused by the treatment of polymorphic
methods like Nil(), where the method arguments alone are not
sufficient to determine a type parameter instantiation. Prompted by
this observaton, I have been working on a new scheme for type
parameter inference which does not share that problem.
Posted to theory by Ehud Lamm on 1/15/02; 6:41:16 AM
|