On the sendfile() stuff, I ahve always regarded this is a bit of a
hack/workaround for not having zero copy IO, ie if all IO is properly COWd
then when the read of the file data into main memory goes to the IP stack
there should be no copies involved (apart from into mbufs, which I don't
think can be helped with the present IP stack). This is one of the things
that I thought Chuck S's merged cache would eventually s9olve (given that
Chuck C's VM had zero-copy IO via a variety of mechanisms as being one of
its defining reasons to exist, I think).

	Neil