Semantic hairball, y'all

I'm in San Jose and the Semantic Technology Conference 2006 has just wrapped up. A good time, as always, and very well attended (way up from even last year. This is an extraordinarily well organized conference). But I did want to throw up one impression I got from one of the first talks I went to.

The talk discussed an effort in "convergence" of MDA/UML, RDF/OWL, Web Services and Topic Maps. Apparently all the big committees are involved, from OMG, W3C, ISO, etc. Having been an enthusiastic early adopter in the first three technologies, I was violently struck by the casually side-stepped enormousness of this undertaking. In my view, all four projects had promising roots and were all eventually buried under the weight of their own complexity. And yet the convergence effort that's being touted seems little more sophisticated than balling all these behemoths together. I wonder what's the purpose. I can't imagine the result will be greater adoption for these technologies taken together. Many potential users already ignore them because of the barrier of impenetrable mumbo-jumbo. I can't imagine there would be much cross-pollination within these technologies because without brutal simplification and profiling model mismatches would make it impractical for an application to efficiently cross the bridge from one semantic modeling technology to the other.

I came to this conference to talk about how Microformats might present a slender opportunity for semantic folks to harness the volume of raw material being generated in the Web 2.0 craze. The trade-off is that the Web 2.0 craze produces a huge amount of crap metadata, and someone will have to clean up the mess in the resulting RDF models even if GRDDL is ever deployed widely enough to generate models worth the effort. And let's not even start on the inevitable meltdown of "folksonomies" (I predict formation of a black hole of fundamental crapitational force). I replaced my previous year's talk about how managers of controlled information systems could harness XML schemata for semantic transparency. I think next year I should go back to that. It's quite practical, as I've determined in my consulting experience. I'm not sure hitching information pipelines to Web 2.0 is the least bit practical.

I'm struck by the appearance of two extremes in popular fields of distributed information management (and all you Semantic Technology pooh-pooh-ers would be gob-smacked if you had any idea how deadly seriously Big Business is taking this stuff: it's popular in terms of dollars and cents, even if it's not the gleam in your favorite blogger's eye). On one hand we have the Daedalos committee fastening labyrinth to labyrinth. On the other hand we have the tower of Web 2.0 Babel. We need a mob in the middle to burn 80% of the AI-one-more-time-for-your-mind-magic off of RDF, 80% of the chicago-cluster-consultant-diesel off of MDA, 80% of the toolkit-vendor-flypaper off of Web services. Once the ashes clear, we need folks to build lightweight tools that actually would help with extracting value from distributed information systems without scaring off the non-Ph.D.s. I still think XML is the key, and that XML schema systems should have been addressing semantic transparency from the start, rather than getting tied up in static typing bondage and discipline.

I have no idea whether I can do anything about the cluster-fuck besides ranting, but I'll be squeezing neurons hard until XTech, which does have the eminent advantage of being an in-person meeting of the semantic, XML and Web 2.0 crowds.

Let's dance in Amsterdam, potnas.

See also:

[Uche Ogbuji]

via Copia
5 responses
+1 to losing the I-one-more-time-for-your-mind-magic  chicago-cluster-consultant-diesel and toolkit-vendor-flypaper -grin-.

But "semantic transparency" in "XML Schema" sounds just a little too scary for my tastes. I'm now regretting not submitting a paper to XTech.. but hey-ho!
Uche, When you get back home, ping me.  You will be happy to see what you see :) (Think AtomicXML + a non-OO SmallTalk + XMPP + Lightweight AspectXML) ... and before you start running the other direction from the description alone, when you see it all in action, it fits together a lot more easily than the above project definitions suggest... its been a break through week, lets put it that way :)
>lightweight tools that actually would help with >extracting value from distributed information >systems without scaring off the non-Ph.D.s.

I recently discovered RDF Gateway, which is fun and productive despite being Windows-only and the ASP-like nature of its apps.

And by the way, my daughter loves your picture.


Fair enough.  I'l start off by giving the terminology some thought.


RDF Gateway has always looked interesting, but it seems more of a DBMS in itself than a broker.  Not really what I was calling for.  There are already a lot of RDF apps, but they're meant for the RDF faithful, for the most part.
Well, it can retrieve RDF (and other) data from multiple sources and implement rules to do inferencing with them, so that takes it pretty far. And, it lets you build apps with reasonably slick interfaces that will be useful to people who don't know or care about RDF/OWL/SemWeb issues, which is a big plus.