W3C home > Mailing lists > Public > public-sws-ig@w3.org > March 2004

Re: A Question about Agent-based Service Discovery/Selection

From: Drew McDermott <drew.mcdermott@yale.edu>
Date: Tue, 9 Mar 2004 19:17:25 -0500 (EST)
Message-Id: <200403100017.i2A0HPj14090@pantheon-po04.its.yale.edu>
To: public-sws-ig@w3.org

> [Gregory Huczynski]
> Basically, in order for a service to be discovered, selected and used 
> automatically, do we not require an initial search-space constraining 
> step based on the notion of known / previously-defined service function 
> interfaces?

Not necessarily, although that is likely to be the case in the near

The alternative vision goes like this: Every web service publishes its
interface in terms that connect low-level message-passing actions to
high-level commercial events.
Example: A web service would publish the (or a) formal equivalent of

     "Send us $15 and we will send you a T-shirt." 

This gets expressed as the predicate-calculus version of

    If you send us a message as specified in WSDL document xxx,
    operation xxx, containing a credit-card number CCN in field xxx
    of data item xxx, and CC is the card with number CCN, and you are
    authorized to use CC, and CC has a remaining spending limit > $15,
    then you thereby authorize us to charge $15 to the account CC, and
    we thereby incur an obligation to give you a T-shirt of type xxx,
    which we hereby promise to discharge by sending you one in the
    next 5 business days.

I claim that most of this is unproblematic, once the vocabulary is
worked out.  It's probably not difficult for the world to agree on the
terms used to represent obligating and authorizing, but the part which
still needs to be fleshed out is (in this case) describing T-shirts.
The problem is that the buying agent may not have used the same
vocabulary as the selling agent.  E.g., the buyer is using a
vocabulary for "rock-band memorabilia," in which one can described
many different things, including the latest Weezer world-tour
T-shirt.  The seller may use a vocabulary appropriate for the garment
trade.  They cover a nonempty intersection (i.e., clothing associated
with rock bands), but there's no guarantee that they used the same
terms for things in that intersection.

In other words, the only things between the vision and reality are the
development of vocabularies for describing many, many domains of
interest to buyers and sellers; and the development of effective means
of translating among them when necessary.

Assuming we can solve these problems (and I could cite thousands of
grant proposals claiming to have solutions), then an agent can work
out the details of connecting to a web services _after_ identifying it
as potentially useful.

                                   -- Drew McDermott
                                      Yale Computer Science Department
Received on Tuesday, 9 March 2004 19:17:26 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:54:12 UTC