Perhaps this is heresy, but I think TBL is overrated. What made the initial web work was simplicity and loose coupling. A lot of what the W3C has done since then seems to be adding layers of complexity where simplicity is best.
And the Semantic Web has been talked about _forever_. The money quote in this article, at least for me, is:
"The challenge, experts say, is in finding a way to represent all data so that when it is connected to the web, links to other relevant information can be recognised and established."
I agree. I wouldn't be surprised to see RDF creep into a niche of the web where hard-semantic data is already a requirement (and this would just be standardizing it), but I feel like in both psychology and web structure, meaning is something that is an emergent property, not something that's hard-coded into the raw data. Determining meaning from chaos is the impressive feat of modern search and the human psyche.
You could say that this is what Microformats are trying to achieve. The problem that Microformats has at the moment is that there isn't enough developers implementing them to make anything particularly useful, which in turn puts off other developers from implementing them as they think "what's the point?". Hopefully in time, Microformats reach a tipping point whereby it's the next hip thing a la Ajax and Rails.
Actually, what I'd like to see is web-wide tagging and then let the tags evolve. For example, I'd like to be able to wrap anything in a web page with <tag name="?"></tag>
The main difference between your idea and eRDF or RDFa is lack of namespaces. The "web-wide tagging" and "let the tagging evolve" bits are a huge part of RDF.
> meaning is something that is an emergent property, not something that's hard-coded into the raw data.
Unless you store everything in unstructured text, it is hard-coded into the data. When you stick an article in a database, you're saying "this is a title", "this is who wrote it", etc.
Sure. A lot of stuff is structured and some standardization may happen there. But let's say that somebody posts a really good comment here that becomes an authoritative source on something or another. You're not going to get the sort of fine-grained meta data that the semantic web expects. Titles are boring. We can get that from HTML already. And web search got a lot better when that and a lot of other structural metadata started being almost wholesale ignored.
Where I could see the semantic web catching on would be in things like Amazon shops or something, where it was advantageous to be able to provide a homogenous interface to a given set of structured data. On the other hand, it wouldn't be in Amazon's interest for that to be the case, because for them the world works better when they get to define the rules and the API and it doesn't easily transfer to another store.
> You're not going to get the sort of fine-grained meta data that the semantic web expects.
Like what?
Articles was a bad choice of example for me to use, because it's a common misconception that the Semantic Web is about making it easier to find documents. That's part of it, but (IMO) data about things is a lot more interesting than data about documents.
"The challenge, experts say, is in finding a way to represent all data so that when it is connected to the web, links to other relevant information can be recognised and established."
That's not the challenge, RDF has made that possible for ages. (You don't even need RDF for that, globally unique identifiers are the important thing. RDF just makes it easier.)
The challenge is getting people with interesting data to publish it in a machine-readable format.
Yes, he clearly doesn't understand the concept that generalizations are not always better than specializations. You can also see this in his old mailing-list exchange with Marc Andreessen on the subject of embedding images (or arbitrary documents including other HTML, as TBL wanted) in HTML.
Bill Gates said the same about Microsoft ten years ago. The only question is when. Apparentely though (hopefuly or not) the money they have will be buying time for them...
The only way I see Google being "superseded" is if they cling to their business model to the extent that it jeopardizes their products. Web advertising, in its current form anyway, already looks like a sinking ship. Unlike IBM and Microsoft, Google inherently has much less user lock-in, so it'll be interesting to see what they do if their cash cow begins to seriously dry up.
Actually, I don't think adoption is particularly important. Really all we need is enough context to probabilistically figure out semantic meaning, and any metadata helps with that. It's not like you need some top-down data schema. Google is already doing it this way already. I don't think semantic web technologies will do much more than make their work a little easier/more accurate.
And the Semantic Web has been talked about _forever_. The money quote in this article, at least for me, is:
"The challenge, experts say, is in finding a way to represent all data so that when it is connected to the web, links to other relevant information can be recognised and established."
Uh huh. I won't be holding my breath.