[MLton] bug in Socket structure signture

[Henry Cejtin]

I agree that the fix is probably just a signature change.  The thing that has
me slightly worried is that I can't find any explicit statement that the
address family is necessarily the same as that of the socket.  Even for
accept, the wording is such that it could mean only that the address family
matches that of the socket returned, not necessarily that of the socket passed
in as an argument.

I can't imagine how that could be, and either way the basis signature is wrong,
so it clearly is better to do the change you checked in.  I'll send Reppy
an E-mail.