> I considered that for about 30 seconds ... and then realized it
> assumes a correlation between mbuf size and MSS/MTU, which is not
> necessarily correct, especially in the presence of different transport
> layers and different networking protocols.

Hm.  Is that really true?  I think that what matters is whether or not
the packet is small with respect to the mbuf cluster size, not whether
or not it's small with respect to the MSS.   Granted, if the MSS is
small with respect to the mbuf cluster size, you have another problem,
but I don't really see how that changes things - if it's a problem to
get small packets, it's a problem to get small packets, and it doesn't
matter whether the reason you get them is because the MSS is small or
because the sender has set TCP_NODELAY.

			       _MelloN_