Google have made an API which they call the Social Graph API. I think the very phrase social graph is bullshit, so I won't say that any more. From now on I'm calling it the social network data API. Stupid name aside, it's a pretty good product.
I haven't bought into the OpenSocial hype. Too much rubbish about Facebook-style apps and widgets. Not interesting. It made all the investors and VCs cream themselves in an orgy, and all the business bloggers go on and on about Facebook vs. OpenSocial, as if that actually mattered in the long run (the social network fight is doomed to one winner - the Web - and that's harder to monetise than closed off stuff, the portals of the current web).
What Google's product does is quite simple. You provide it with a URI and it'll return a list of linked resources - either other facets of your own identity (other pages you control, other social network profiles), connected friends and contacts or both. It uses a combination of XFN and FOAF.
The XFN performance is stunning. If you put in details of people who are big users of XFN (Tantek, Kevin, Jeremy etc.), you get tremendously accurate and useful results. In the case of XFN, it functions much as Plaxo's Open Social Graph tool, only more useful and a lot quicker (it's not probing each page - it's Google, remember).
Now, for FOAF, it's less accurate in a number of ways. It's not parsing FOAF in the way that most folks probably would. I haven't quite pegged the problem yet, but there's something slightly weird going on with it. We've been discussing it on swig (logs: yesterday and this morning). And Danny is quite concerned about the SGN pseudo-URL scheme that Google are using, and says it's breaking some fundamentals of Web Architecture (see his comments on Brad Fitzpatrick's blog post).
To be honest, I'm not too bothered about the SGN thing. You can specify in the API call whether or not you want SGNs or proper URLs. Yes, it may technically be a breach of Web architecture, but it's not one that really concerns me too much. It seems slightly silly when we already have a method in FOAF of specifying account names (the OnlineAccount class and it's subclasses, combined with the holdsAccount, accountServiceHomepage and accountName properties - I use all of these in my FOAF document to describe Facebook friends who don't have URIs but have unique numbers), but it's not the end of the world.
What's the new Google tool going to do then? Well, some people are going to use it for the fairly obvious first use - letting you import friends from other sites and prove one's identity etc.
This evening I've been figuring out how we get the data back out again and on to the Linked Web of Data. I've been prototyping an RDFizer. It's pretty simple to do - nothing that is of any real difficulty. I'm using Python to prototype it, but hope to have it available in a few different languages very soon (Python is a given, PHP is too - maybe Ruby, and Java if I feel masochistic - and most definitely available as a web service). Okay, so we get the data out as RDF, what then? Well, there are two things I can see which would be really cool. One is desktop caching. This is not hard. You could have a really neat little application that uses, say, SQLite and librdf as a little triple store. That way, when you run across a page on the Web, you could have a little sidebar or pop-up in your browser that would show you "this person knows five people whose web pages you've browsed recently". Okay, you could do that without the RDF and triple-store, but it's a cool use case nonetheless.
A slightly more practical and useful point is that we can use it as a scutter plan in order to pull in data about our whole social network. The combination of the Google Data plus some shareable rulesets and scripts could be a useful scutter plan. It'll be interesting to see how else this can be used as part of Linked Open Data.
What else, then? I'm pretty sure that a simple mechanism for XFN 'completeness' will mean that OpenID whitelisting for expanding hCards can become more popular, as I described back on 2007-11-05. Mine is still based on FOAF though - just because we haven't got the full stack together for XFN (it's almost there, and one could do it, but FOAF for me already does everything that's needed).
I've gotta say, SGNs aside, a big thank you to the folks at Google for making a big contribution to making XFN and, to a lesser extent, FOAF more useful. This is in the long-run more interesting than all the vampires and zombies crap on Facebook put together.
