Simon Marlow writes ("RE: [Socket] accept & recvFrom"):
...
> You're suggesting removing [Socket.recvFrom]? Again, I've no objection (because
> I don't use it :-) but it's harder to make a case for removing something
> from a library. In this case, recvFrom is the dual of sendTo, so it at
> least looks consistent. I agree it's hard to imagine a situation where you
> might use it.
Speaking as a UNIX networking type of person, I too think that the
sendTo function is, well, rather weird, I have to say :-). That
doesn't mean I think it should be removed, but I can't imagine using
it myself.
However, there is the potential objection to both recvFrom and sendTo
in Socket that they have name clashes with the functions sendTo and
recvFrom in SocketPrim, which do exactly what I would have expected
(and are essential for use of datagrams).
Also, of course, stupid people might try to use sendTo and recvFrom to
have a communication with (eg) an SMTP server, but hopefully such
people will not be able to figure out Haskell well enough to get that
far :-).
Ian.
