x-kernel

[John Jannotti]

It's very hard to say that their design is too slow.  I mean, sure, it's
slow, but it's on a Sun 3.  Like I said, I suspect their design is too slow, 
since they associate a thread of execution with each packet, but then again 
they seem to have beat a lot of the other work at the time.

My point was not that they came up with micro-protocols, just that they
began to realize that their boxes could be used to "implement the IFs" of
protocol processing, like a language.

Our strength is our model: explicit queues, push/pull, and the flexibility
of using many factions to make up a single protocol (and we show that this
is efficient, they only speculate that it's possible).  But it is
interesting that we could practically steal entire paragraphs from the
"Discussion" section of their paper and have them fit it smoothly with
ours.  They talk about the modularity imposed by this architecture, the
uniformity of interface between nodes, and reusability of micro-protocols
if they are broken into separate nodes.

  jj