Paper Title: A Datastream Language for Protocols
Letter grade: B-
Numeric grade:
Confidence
Comments:   Good paper,

 It wasn't clear who the audience was for these protocols.  Are they intended
for protocol designers working in the operating system, or for application
designers who would like to try ILP and ALF?  The paper is interesting, and
should be in the workshop, but I believe that the authors must deal with two
concerns before they get to the workshop:

	- the paper is motivated by performance, but there is no data to yet
indicate that this is a good idea. In fact, intuition suggests that this
approach will result in protocols that run more slowly than conventionally
structured protocols. The authors may have some insight into why this intuition
is not applicable, but they've not done a good job of presenting that insight.

	- the authors don't make a convincing case that something real, beyond
very simple protocols, could be built.

The other paper on automatic protocol construction (header prediction) is
substantially stronger than this one, so if there are limited slots, the other
one should be taken.