[E-Lang] Performance and complete mediation

[Jonathan S. Shapiro]

> Our measurements showed that buffer copy overhead started to show up at
> around 1 KB messages.  By 500 KB, it dominated.

I agree, though if you get the context switch built right it dominates even
sooner -- around 1Kbytes.

However, it's important to acknowledge that the copy overhead is a
*necessary* cost. There is a certain minimum time to move the bytes, and a
certain minimal overhead to validate the mappings, and you really wouldn't
want to give up either of these operations. The question we want to ask,
then, is given that there is a hard minimum to the cost of
protection-motivated copies, is this cost acceptable in the context of some
given system.

Jonathan