Dave,
However, I can see a strong case for restricting the top level rule
language to only fire off header information and leave it up to the proxylet
to implement any body matching rules as part of its internal implementation.
That's what we had in mind, yes.
This would be particularly powerful if the proxylet is free to add new
pseudo-header information which can in turn trigger other proxylets.
Yes, that's a good point, we also discussed this earlier. IRML will
allow to have user defined headers in the property element.
The example you mentioned also is a nice example in which the rule
engine has to operate on modified headers rather than always on the
original ones. If the first rule matches and the associated action
adds one of the user defined headers you mentioned, the next rule
match should be done against the modified header, and NOT the original
one.
-Markus