tuple flattening

[Suresh Jagannathan]

I've (re)started work on the opus, and in particular the tuple-flattening
section.  After attending POPL, I was inclined to believe that there might be
enough to warrant submitting the flattening specification (along with specific
incarnations) to POPL this July.

However, in reading the translation we currently have, I'm wondering how
significantly different it is from Leroy's 1992 POPL paper on unboxing objects
and polymorphic typing.  In that paper, he suggests introducing coercions at
function call and return to specialize and generalize boxing operations.  Isn't
that essentially what we're doing here?  Of course, his type system has to deal
with source level polymorphism, while we don't, but the solution he suggests
seems pretty close to ours.  Do you agree?

      -- sj