Joshua Allen responds to Marc Andreesen's types of platform.

He says the new platforms are about "data". That's rather like Tim OReilly's "Data Inside" aspect of his web 2.0 definition.

But a) it's not the whole story. And b) it leads to demands for "opening" the platform by making the data migratable.

It's not the whole story because, in addition to data the's user, her social connections and social conventions are the platform. When you think about it, this is true even with Windows, which is why UI consistency is such an important part of a platform.

On social platforms, the lock-in comes not from just having the data walled up in your silo. It also comes from your network being the place where people tend to do X. And if, on somebody else's platform, people don't tend to do X, then they won't shift the X-related applications over.

Because of his "open-data" perspective, Allen, I think, under-estimates the importance of the hosting issue, although he understands it perfectly well.

It's obvious what benefit a Ning or Salesforce.com would get from keeping your data and code on their servers; it's less obvious what the benefit to you is. There are only two real reasons such an arrangement would be a benefit for you :

If your data, aggregated with data from lots of other customers of the provider, can provide some additional intelligence.

If the provider gets dramatic economies of scale beyond what you could get on your own. In the case of a Ning or a Salesforce.com, this one is dubious. There are only a handful of companies who buy electricity and bandwidth in enough volume to offer hosting cheaper than Amazon. Companies like Yahoo!, Google, and Microsoft.

This is the same debate that's had around Software-as-a-service. It's that "additional intelligence" which is the killer thing that social platforms can offer : doing stuff with social data that cuts across users and across applications. That's why Amazon's database is so much more valuable than everyone hosting their own "I bought and like this book". It can figure out the most popular or "readers who bought X also bought Y".

Yes, theoretically, "scutter" applications can run around a widely distributed microformats, but my bet is that the difference in efficiency and difference in privacy control is actually so great, quantitatively, that it can lead to qualitative differences of application.