Linux 2.4

[Stephen Weeks]

> Mind you, I'm surprised that this is really allowed.  Is it just one  of  the
> un-specified parts of the basis library?  I would think that this would cause
> all kinds of programs to fail to type check.

It causes programs to be very nonportable.  This sucks and is a stupid decision
on their part.  I mentioned this to Dave MacQueen a month ago when he was here
and he agreed, but did not think that Reppy et al. had fixed the problem in the
ever impending spec.

> I would think that once this is understood, it would  only  be  a  couple  of
> lines of change to MLton.

Int64 seems easier than IntInf to me.  At some point MLton would have to convert
to 64 bits to call lseek, so we would have to add support for Int64 no matter
what Position.int is.  Unless it's unsigned 64, in which case I guess we have to 
make it IntInf since SML has no notion of unsigneds.