Semantic Web Interest Group
I'm spending this week -- in my role as a researcher at the University of Maryland -- in Cannes-Mandelieu, France, at the W3C's 2004 Technical Plenary. Yes, Cannes is as beautiful as you might think. In fact, it's more beautiful than I imagined; it reminds me, somehow, of a cross between Seattle and Phoenix. Weird, I know, but trust me: you should plan a visit to the French Riviera if at all possible.
So what is a Technical Plenary (TP)? It's a series or cluster of meetings, mostly of W3C Working Groups, together with other W3C subpart meetings. And there is a plenary, or all-hands meeting, during which various presentations are made and discussions are had regarding issues facing the W3C and its membership. The other thing to say about this TP is that, oddly enough, while I'm presently in the TP, I'm not officially allowed to tell you about it, since the membership voted to make it a non-public event. I can tell you that I'm at the TP, that it's interesting, and that many people are working hard, but that's about it.
I suppose I can also report that the W3C's TAG had a face-to-face meeting, at which I was an observer, and during which I observed discussion of various Last Call comments.
The good news, however, is that the Semantic Web Interest Group (SWIG) has been meeting in conjunction with the TP, and the SWIG proceedings are public.
Semantic Web Interest Group
The Semantic Web Interest Group, as the successor group of the RDF Interest Group, was "designed as a forum to support developers and users of Semantic Web technologies such as RDF and OWL". Dan Brickley, as chair of the SWIG, chaired the two day meeting, which included presentations, lightning talks, and open discussions.
The most exciting aspect of the SWIG meetings, in my view, is that it was very large, very robustly attended, and full of developers who are new to the Semantic Web and its technologies. Yes, of course, there is a stable core to the Semantic Web community; that is, no matter which conference I attend, on which continent, I see many of the same faces repeatedly. But it's also important that there is a mixture of new and old, as well as a constancy of churn or turnover between the new and the old. I am encouraged by the vitality of the Semantic Web community as evidenced at the SWIG meeting in Cannes.
Dan Brickley spent the first bit of time on Monday introducing people to the various social forms of interaction and process in the SW community, including mailing lists, discussion forums, and overviews of W3C history and process.
Mark Birbeck, author of a recent W3C Note called "XHTML and RDF", discussed his proposal for solving the old problem of embedding RDF in valid HTML or XHTML. I expect XML.com will be publishing an extended examination of Birbeck's proposal; but let me say now that it seems to have the favor of the relevant Working Groups. It will be good to see this problem finally solved, at least from the perspective of specifications.
Graham Klyne has been working on Swish (Semantic Web Inference Scripting in Haskell). Klyne has also been pushing, ever so politely, in the Haskell community for it to take web services and the Semantic Web seriously. I appreciate that effort as Haskell is the language I most would like to learn. With Swish one can do inferencing on RDF graphs, with the goal of supporting Haskell as a kind of "scripting language for the Semantic Web". Klyne spent some of his time talking about some of his work supporting datatyping in RDF and OWL.
Patrick Stickler, who works for the developer support arm of Nokia, described his ideas for a Semantic Web Server, including his controversial suggestion to add new methods to HTTP -- MGET, MPUT, MDELETE -- and his notion of a "concise bounded resource description". Stickler and Jeremy Carroll also talked about their new proposal for an XML serialization of RDF, which they call TriX -- a serialization which is a very triples-oriented, records-like view of an RDF graph.
José Kahan, a W3C Team member who's worked on Amaya, presented his ideas about bookmark management using RDF. In order to solve the problem using relative URIs in bookmark files, Kahan has decided to use URNs which are mapped to URIs -- a solution, I realized earlier this week, which is similar to parts of the Atom specification.
Mike Uschold, of Boeing Corporation, described his use of OWL and description logics to filter messages in a publish-subscribe system used by battlefield commanders. Typically such message filter expressivities are based on path-like languages, often XPath. Uschold's work is interesting because it shows that the increased expressivity of a description logic or ontological approach is useful in many different contexts, including ones which aren't necessarily being used on the Web.
Uschold's work also suggests that OWL and description logic processing may end up having a role to play in the syndication of news and other resources on the Web. There are several sites which presently offer a pub-sub notifications of various kinds of events, as represented by changes in RSS feeds. If this technology is to mature, I suspect that users will eventually clamor filtering tools which are more expressive than either keyword matches, XPath queries, or regular expressions. A real-time aggregator over semantically rich data, using complex ontologies as the filtering technology, is not that far away and could be a very powerful tool for building lenses or views onto, for example, slices of the Web like weblogs.
Finally, though I have only skimmed the surface of the two day SWIG proceedings -- including a series of very informative lightning talks, about which more can be found in the RDFIG logs -- I want to mention Adobe and its XMP project. Adobe, as you probably know, has embraced RDF for digital resource metadata in a big way. Just about every tool in the Adobe arsenal is now able to embed RDF into their primary artifacts. It's ironic, of course, that a core web technology, RDF, still isn't really embeddable into the core web data format, HTML; but it can easily be embedded into Adobe artifacts.
If my prediction of a pub-sub notification system, using OWL as the filtering technlogy, seems a long way off, consider what Chuck Myers, the Adobe representative, said about XMP. Myers suggested that there are two reasons why Adobe customers want metadata embedded in digital artifacts. The first, obvious, reason is to support automated workflow. If you've ever worked in a graphics shop or design house, you know that workflow management is absolutely critical to profitability. The second, less obvious, reason is intelligent syndication. What a real-time OWL-powered pub-sub application needs, of course, is rich RDF metadata attached or linked to artifacts of the Web: images, video and sound files, as well as XHTML documents. Two of the missing pieces of this puzzle, how to embed RDF into XHTML in a valid way and how to empower content creators to annotate the stuff they make, appear to have been solved and are on the way to being deployed. Having Adobe, the dominant force together with Apple among the creatives, behind RDF and the Semantic Web will go a long toward achieving success.
D, O, Double G
Finally, I want to mention the newly formed Data Access Working Group (DAWG), which comes under the auspices of the Semantic Web Activity. Co-chaired by Dan Connolly -- who's looking for another co-chair, apparently -- this WG is chartered to work on two issues: first, creating a standard for an RDF query language; second, creating a standard data access protocol -- likely via SOAP and plain HTTP (REST) -- for interacting with RDF data servers.
There are something in the order of twenty different implementations of RDF query languages, many of which are path languages for navigating through RDF graphs. And for every triple store that has any sort of client-server support, each one does things differently enough that interoperability becomes a headache. Unlike some other parts of the Semantic Web Activity, the work of DAWG will be almost purely an effort to standardize a technology and solution space, rather than doing any kind of research.
Also in XML-Deviant
This focus on standardization instead of research is an exciting turn of events. First, it's exciting because it means that DAWG will pursue an ambitious schedule. I have heard loose talk of a standard within six months.
Second, it's exciting because it means that we could soon, inside of 12 to 15 months in my estimation, have widely deployed libraries for most programming languages that play a role for the Semantic Web analogous to the role HTTP itself plays for the Web proper. In other words, we're at the point now where the basic knoweldge representational mechanisms -- RDF and OWL -- are formalized; where mechanisms for the creation of RDF and OWL are coming online and into place; and where there is increasing awareness of and interest in the SW. What we need, then, is the "uniform means of interaction" for the SW that HTTP provides for the Web. What we need is a reasonable specification that we can give both to RDF tool makers and to to Python and Perl and Java and C# and Ruby programmers; and we need to say to them, "implement this specification in your tool or your language; then your users and programmers will be able to uniformly access resource representations on the Semantic Web". That will be a happy day, indeed.
Third, and most personally, it's exciting because I'll be a member of the DAWG in my capacity as a UMD researcher. I count myself fortunate that my first Working Group will be one which is responding to real needs and offering real solutions to a user and developer community that is poised to do big things.