- From: Dan Brickley <danbri@w3.org>
- Date: Mon, 10 Nov 2003 17:38:32 -0500
- To: clay@shirky.com
- Cc: em@w3.org, www-archive@w3.org
Hi Clay, I've been following the discussion around your recent piece on the Semantic Web. I dropped a few comments into weblogs here and there([1]) since your site doesn't have a comments facility, but thought I'd send you a quick note too. It's always interesting to see how others perceive these things. The main thing that struck me was how the logical rules / formal inference side of Semantic Web dominates your perception of what we're trying to achieve with RDF and related technology at W3C. You're certainly not alone in that; many of the KR/AI crowd have got involved, and several of my colleagues here at W3C certainly find that aspect of the technology quite compelling. However if you look at the work we've been doing, particularly around the core RDF specs[2] I think you'll find much of it reassuring near-term in scope. Since your piece is couched explicitly in terms of W3C's work, I was dissapointed that you missed out on this. Many RDF apps get by perfectly well without any fancy inference rule machinery, exploiting the RDF data model as a handy mechanism for mixing independently created data vocabularies. The Semantic Web project, viewed as an effort to make it easier to publish, mix, share and consume data on the Web, depends on logic in pretty much the same way SQL or UML depend on logic. Creating an RDF vocabulary (ontology, etc) involves making a number of pragmatic tradeoffs, between predictability (nice tidy categories) versus real world messyness. These tradeoffs are familiar to database designers the world over, and not peculiar to the SW effort. One point I was particularly puzzled by -- perhaps you could help me out here -- was your apparent impression that we're working towards a single, global, monolithic ontology. The reality of SW hackerdom that I'm familiar with is quite the opposite. There are a number of evolving RDF vocabularies, each of which describes some part of the puzzle in a level of detail, and from a perspective, appropriate to the needs of those creating it. If it works for others too, in part of whole, that's great. If not, there's plenty of room for another perspective in the SW vocabulary marketplace([3]). In this regard, the SW to me feels reassuringly bottom-up, grassroots led (eg [4]); a characteristic that stems from the technical design of RDF. Unlike vanilla XML, RDF vocabularies can be freely mixed together in data without prior agreement. So you often see ad-hoc combinations of Dublin Core, RSS1, MusicBrainz, RDF-calendar, FOAF, Wordnet, thesaurus, Geo-info etc etc frequently deployed together, despite the fact that the creators of those various vocabularies barely knew each other. This strikes me as the height of loosly-coupled pragmatism rather than a wide-eyed effort to build a monolithic universal category system. But if this is they way our project looks from the outside, it is good to know! I had a few more comments in [1] regarding the Friendster mention towards the end of your article. I think you missed a point here. SW technology allows, but by no means requires, people to make simplistic over-generalisations. My own experience with FOAF is that lots of people do want to exchange data that embodies sweeping and simplistic generalisations about subtle social relationships. But you can make such mistakes on any platform (see various SQL-backed social networking sites for details). SW technology, specifically RDF, makes it *possible* to goof up in various ways, but it also allows for subtler treatments, which is where (hopefully) FOAF is headed through its focus on describing the photos, events, collaborations etc that are the evidence friendship leaves in the world, rather than crudely taxonomising classes of friend. So, again, it felt like a strawman was being punished in those observations. From where I'm stood, the SW project is all about getting RDF descriptions of things deployed in the public Web. The vocabularies that support those descriptions (ontologies) come from a variety of sources, adopt varying tradeoffs, and -- inevitably -- overlap in scope and coverage. What we have created in RDF, I believe, is a framework for allowing these ways of describing the world to compete in a way that lets application developers adopt a fine-grained approach to using them. This is a technical characteristic of RDF which counts against the 'global ontology' concern you raise: RDF's design makes it easier to pick'n'mix pragmatically from various pre-existing vocabularies, adding in extensions and qualifications of your own where needed. Hmm OK this was supposed to be a short reply! I'll copy it to www-archive@W3.org so it'll show up in http://lists.w3.org/Archives/Public/www-archive/ -- if you've time to respond, copying www-archive will similarly put your reply in Google-able space. (Short version of this might've been: there's more to the SW than rules and inference...) cheers, Dan [1] http://dannyayers.com/archives/002016.html#comments [2] http://www.w3.org/RDF/ [3] http://esw.w3.org/topic/VocabularyMarket [4] http://www.xml.com/pub/a/2001/07/25/rdfcalendar.html
Received on Monday, 10 November 2003 17:38:52 UTC