Comparison between plan MS and MC

Plan MC (type functions in GHC)

* local assumptions are turned into rewrite rules

* "simple", easily fits into GHC's current scheme

* but has some restrictions, eg.

• (local) equations must be oriented

• (local) equaitons must be terminating and confluent

• only during constraint generation we may be able to test whether conditions are satisfied

Plan MS

* maps the problem to CHRs

* more complete, less restrictive but may require more substantial changes to GHC's inference engine