Re: questions about Process.owl

From: Drew McDermott <drew.mcdermott@yale.edu>
Date: Sat Mar 04 2006 - 16:51:33 EST

> [David Martin]
>
> I'd also like to add a new question of my own, for you, Drew: Do we
> need to have both Produce and Write, and what's the distinction
> between them?

The distinction currently is that Produce sets an Output and Write
sets a Link. It's a very interesting question whether these can be
identified. Outputs are "heavy-duty" things that (may) get grounded
if this process is used to model communication with an actual Web
agent. Links are for interstep communication. But what the hell,
BPEL unifies them (I think) under the name "link." I've sort of been
thinking that, just as links can be used to synchronize inter-step
communication, outputs are used to synchronize inter-agent
communication. Identifying the two makes a lot of sense.

The way to do it is to make Output be a subclass of Link, replace
OutputBinding with LinkBinding, make the argument of Produce be a
LinkBinding, and do away with Write. I don't think this will screw up
the grounding ontology, because Outputs remain special in that they
are the subjects of grounding. Grounding.owl refers to
&process;#Output, but doesn't appear to refer to OutputBindings.
In future it might be possible to talk about "internal groundings" of
Links. (Note that Locs, the general-purpose variables, are pretty
much useless for thinking about inter-step synchronization, message
passing, grounding, etc.)

I'll go ahead and do this, but if someone wants to pull the plug, we
can do it in the interest of conservatism. Actually, I think making
the inside of a process look more like the outside is something
Bijan has always argued for, so if he's listening I wouldn't
anticipate an objection from his zone. Anyone else?

                                             -- Drew

--
[To unsubscribe to this list send an email to "majdart@bbn.com"
with the following text in the BODY of the message "unsubscribe daml-process"]
Received on Sat Mar 4 16:49:34 2006

This archive was generated by hypermail 2.1.8 : Sat Mar 04 2006 - 16:49:35 EST