W3C home > Mailing lists > Public > www-archive@w3.org > November 2001

concat: the missing link from SW to WS

From: Dan Connolly <connolly@w3.org>
Date: Wed, 28 Nov 2001 02:43:40 -0600
Message-ID: <3C04A3BC.97F715BF@w3.org>
To: www-archive@w3.org, hugo@w3.org, danbri@w3.org
A while back, I said that if TimBL would just
add string concatenation, I could use
that along with log:uri and log:resolvesTo (now
called log:semantics) to do semantic web services.

TimBL has since added string concatenation,
but I didn't follow up right away on the
web services angle. [where's that agent
that does notification when pending dependencies
are satisfied? anyway...]

Tonight I just got a use case working: I've
combined a database-backed web service with
semantic web rules to answer the query:

	Who's been nominated to the WG but
	not subscribed to the mailing list?

This is about 1/3rd usefully general stuff
and 2/3rds kludge-for-now stuff...

Given the "group id" key into our database-backed mailing
list subscription system, I can conjure up
the address of a service that provides the
subscription data in RDF; then using log:uri,
I can treat that address as a URI/symbol...

this log:forAll :grp, :doc, :pg, :who, :mbox, :fname, :gname.
this log:forAll :grpNum.

{ :grp dbwg:group :grpNum.
  :pg log:uri [ string:concat
    ("http://www.w3.org:80/2000/09/dbwg/detailsRDF.php3?" "group" "="
  { :grp :subscriptionPg :pg }.

TODO: I'd rather key of the WG home page than
its dbwg:group, but the WG home page isn't in
the database yet.

TODO: Also, I'd like to figure out a richer idiom
for connecting the dbwg:group property
to the ...detailsRDF.php3?group=12312 address...
something to deal with application/x-www-form-urlencoded
in general.

TODO: and there's the whole N3/SQL remote-query
pile of stuff; I haven't gotten very far in my thinking
about that yet...

Meanwhile, then using log:semantics, I can get the actual
data and process it.

In this case, I use log:notIncludes to find
missing subscriptions:


#find missing folks from dbwg database
  :grp org:charter :doc;
    org:member :who.
    contact:mailbox :mbox;
    contact:familyName :fname;
    contact:givenName :gname.

  :grp :subscriptionPg :pg.
  :pg log:semantics [ log:notIncludes {
    [ org:member [ contact:mailbox :mbox ] ].
  } ].

 log:implies {
  :pg :missingSubscription :mbox.


And I'd really like to compute the data that should
be POSTED in order to add the missing subscription.

In this case, it's messy because it might involve
two transactions: (1) creating an account for the
subscriber, and (2) subscribing that account to the group.

But if I had that, I'd like to run it every night
and have it notify me if something's missing;
the notification would have everything all set
up to synchronize, ala:
	"<DOIT>: subscribe <foo@glorp> to the WebOnt WG,
		cuz he was nominated 21 Nov 2001"

For that, we need SemanticWebCron because it
needs access credentials. And to do SemanticWebCron,
we need ProofBasedAuthorization. Turtles all the way down.

Meanwhile, I nearly got completely depressed because
the data are so messy.

there's no logically compelling argument that the guy who
gave oisin.hurley@iona.com as his email address when he
applied for member-access is the guy who was nominated
to the WG with mailbox ohurley@iona.com .

I could add independent assertions that the same guy
uses both those addresses, but (a) I'd have
to use sameThing/forgetDups all the time, which
slows cwm to a crawl, and (b) I can't figure out
how to say "if, for each of the guy's mailboxes,
<detailsRDF> doesn't say that mailbox is subscribed,
raise a flag about that guy". Well.. I can say
it using forAll, but not in a way that cwm will
act on.

I ended up with two rules:
	if there's a (nominated) member with mailbox X
	but no X in the subscription list, conclude
	that the subscription list is missing X.

	meanwhile, if there's a nominated member with mailbox X
	whose surname is N,
	and there's a subscripition with mailbox Y and surname N,
	then note that Y might be an alternative
	to one of the missing mailboxes.

Anyway... tonight I claim victory for gluing
the Semantic Web to Web Services by using
string:concat along with log:uri and log:semantics.

The results at this point:

mm:missingSubscription  [
             :mailbox <mailto:Frederik.Brysse@ivisgroup.com> ],
             :mailbox <mailto:warner.ten.kate@philips.com>;
             mm:maybeSubscribedAs <mailto:warner.ten.kate@philips.com>
             :mailbox <mailto:kohlhase@ags.uni-sb.de>;
             mm:maybeSubscribedAs <mailto:kohlhase@ags.uni-sb.de> ],
             :mailbox <mailto:libby.miller@bristol.ac.uk>;
             mm:maybeSubscribedAs <mailto:libby.miller@bristol.ac.uk> ],
             :mailbox <mailto:ohurley@iona.com>;
             mm:maybeSubscribedAs <mailto:ohurley@iona.com> ],
             :mailbox <mailto:lynne.thompson@unisys.com>;
             mm:maybeSubscribedAs <mailto:lynne.thompson@unisys.com> ],
             :mailbox <mailto:jeremy_carroll@hp.com>;
             mm:maybeSubscribedAs <mailto:jeremy_carroll@hp.com> ],
             :mailbox <mailto:lorbst@mitre.org> ] .

 -- http://www.w3.org/2001/sw/WebOnt/missing-stuff.n3
 built from

(lots of stuff needs chacling; sorry, it's late)

Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Wednesday, 28 November 2001 03:43:45 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:42:03 UTC