DataTweet runs at Amazon EC2, and relies on the Twitter API.

Background

Twitter's approach to the global information stream should apply with equal effectiveness to content from sensors or other automatic agents, rather than from the typing fingers of humans, or so it seemed to me, your host here at DataTweet. I'm as interested in how high the waves are off shore where I live, or the current K-index (a predictor of auroras), as in the latest remarks of friends or big-time tweeters.

Of course, data can be inserted directly into a twitter stream over the twitter API; here are a few examples: http://twitter.com/tweetidor, http://twitter.com/earthquake, http://twitter.com/FirefoxCounter. Also, people can tweet structured data if they follow rules, as proposed at Microsyntax.org. This is useful, but structured data begs to be filtered and organized. It is rare that the whole of an unfiltered data stream is appropriate for direct inclusion in your timeline - almost always, you'd like to follow some portion of interest. That is the core of what DataTweet does: provide the means for following filtered datastreams, and of presenting the result in a way that exploits stucture (currently only via maps but graphs, sorting, etc, are coming).

It is less important where the streams are hosted. In the initial release, DataTweet hosts its own streams, since there are some practical downsides to Twitter as a data-streaming service, at least for now. First there are the rate limits: no more than 1000 updates per day. Second, acquiring and filtering data from Twitter by polling multiple streams is not very efficient, and the acquisition process is subject to its own limits. This may change with their new streaming api (in alpha test) - and depending on interest, this is a reasonable future approach for integration of Twitter-hosted data streams into the DataTweet environment.