A recurring question that us Semantic Web weenies still get is "why do we need RDF? We can do it all in [insert favorite technology here]". That's a trick question, of course. Sure, you can do the same thing in different technologies. One way to respond would be the rather fluffy, "well, you can do it better in RDF".
But today, I'm going to respond to the question a different way.
If you can do it in XML, then why didn't you?
Today's example is geotagging in Flickr. The idea is that if you put a couple of special tags on your Flickr images, that you can do cool things like look for pictures other people took of the places you have visited, find the most popular places for photo shoots, etc.
Back in 2005, some folks decided that this was a good enough idea that they proposed a hack for how to do it, and published how-to guides. If you hunt through Flickr, you will find a number of content providers who have done this.
More recently, Flickr has provided tools for doing this inside of Flickr itself. With these tools you can pinpoint the place where your photos were taken on a map, and have the tags generated for you. Then the Flickr geo aggregators can show you mash-ups of where your photos were taken.
Now let's start publishing this stuff on the web for other people to use. Let's syndicate our Flickr content via RSS or Atom, and use that geotagging data to combine with some other things, like, say, the location of special events, attractions, transit systems, etc. Did anyone happen to be taking pictures near the grassy knoll when Kennedy was shot?
But wait a minute. What's that you say? Flickr doesn't export its geo tags in its feeds? Oh. Sorry to hear that. Never mind. I haven't found an XML developer who thought that the idea of amalamating geo feeds was "too abstract" or "an academic exercise". But it might as well be; Flickr won't give you the data you need to do it.
Next time someone asks me why we need standards for the semantic web, that's the story they'll get. So that we can share the information that Flickr keeps proprietary.
But wait a minute . . . what about those old-fashioned taggers. How did they do it? Wasn't that using Flickr's standard tagging mechanism? Maybe we can export that?
Well, as it happens, the old-fashioned taggers suggested that you create a new tag, of the form "geolat52367057" to describe the latitude of the photo you took. And as it happens, a lot of people did that. I'll leave it to you to figure out where the decimal point goes; but there is a lot of Flickr data with this kind of metadata.
How does it look when we export it as an RSS/Atom feed?
Well, that depends on how you export it. Flickr's RSS 2.0 generator crams all the tags into a space-delimited string. Blast out the ones you want. The Atom generator is a bit more sensible; at least you get one XML tag per Flickr tag; you still have to blast out the "geo" thing. How about RSS 1.0, the RDF-compliant RSS feed? Surely it does better? Well, surely it could, but you can only lead a horse to water. Flickr chose not to drink; the Flickr tags don't appear in their RSS 1.0 feed at all.
So, if you were hoping to federate geo tags from Flickr with your other geo data, you've got a lot of work on your hands. Web 2.0 seems to have almost got the idea of the semantic web, but can't quite make it work. Maybe one day it will.