[Prev][Index][Thread]

clarification of recently-posted abstract






          Standard ML weak polymorphism and imperative constructs
 
               My Hoang   John Mitchell  Ramesh Viswanathan
                     Department of Computer Science
                Stanford University, Stanford, CA 94305
              {hoang, mitchell, vramesh}@cs.stanford.edu


The recently-posted abstract may be somewhat misleading regarding the
difference between Standard ML, as specified in the formal definition,
and Standard ML of New Jersey, a popular implementation of the language.
The typing of imperative constructs in Standard ML is fully specified in
the formal definition [1], and proved correct in [2,3]. It is the SML-NJ 
implementation, which differs from the formal definition, that uses weak 
type variables and has not been analyzed previously. The main differences
between the two approaches are explained in the paper.

Sorry for any confusion on this point.

John Mitchell

[1]     @book{MTH90,
        author = "Robin Milner and Mads Tofte and Robert Harper",
        title = "The Definition of {Standard ML}",
        publisher = "MIT Press",
        year =  "1990"
        }

[2]     @PhdThesis{tofte,
        author = "Mads Tofte",
        title = "Operational Semantics and
        Polymorphic Type Inference",
        school = "Edinburgh University",
        year = "1988",
        note = "Available as Edinburgh University
        Laboratory for Foundations
        of Computer Science Technical Report
        ECS--LFCS--88--54."
}

[3]     @article(Tofte90,
	author="Tofte, M.",
        Title="Type inference for polymorphic references",
        journal="Information and Computation",
        Volume="89",Number="1",Year="1990",pages="1--34")