[MLton] the next public release

[Stephen Weeks]

> -keep-sml doesn't work for .mlb files, although -stop sml does.
> Note that in either case, you probably can't naively send the
> resulting sml through MLton, because the elaboration of mlbs allows
> you to hide and reveal module level defintions.

Yeah.  This should be mentioned in the user guide.

> The major caveat with CML is that the Basis Library is not thread
> safe and there are no CML event versions of the Basis Library IO.
> The "right way" to call a Basis Library function that is stateful is
> with MLton.Thread.atomically.

Fair enough.  This should also be mentioned in the user guide, and
probably also near the top of the README in lib/cml.

> We should probably also provide ml-yacc-lib as an .mlb available
> library.

Sure.

> I also wonder if with the .mlb model, we shouldn't have
> $(SML_LIB)/basis/basis.mlb correspond to the strict 2002 basis.
> That is, to gain access to MLton or SMLofNJ or Unsafe structures,
> you need to reference additional .mlbs.

Definitely.