Activity Streams for the Open Web

We’re seeing this proliferation of life streaming services in the Web 2.0 world. Lifestrea.ms, FriendFeed, Scope and readr just to name a few. These sites work by aggregating user feeds into one combined feed, known as an activity stream, that is then readable by the user who sets it up. With a glance, they can see what all of their friends are up to (assuming their friends are using their streaming services of course). While these activity streams are interesting from a user standpoint, they aren’t very machine friendly. While I can read these streams on a web page for via RSS, getting at the data in these feeds is near impossible.

I’ve been talking with Chris Messina about this for the last couple of weeks and he had a really good post to the DiSo (distributed social networking) mailing list this weekend about it:

One of the things that I think is critical for DiSo to work on is the distribution of activity streams (aka lifestreams or newsfeeds). As Marshall Kirkpatrick called them, these “Standards Based Nerve Centers” or “Open Aggregators” (Dave Recordon) potentially provide the value of DiSo, harvesting activities from across the web, from friends but also from our own actions, and, with some work, can begin to provide some smarts in terms of “accelerating serendipity” — introducing us to new people or to interesting things that we might not otherwise have come upon.

Chris goes on to describe a possible example of marking up an activity stream with microformats (just POSH - plain old semantic HTML):

Turn this:

“Chris listened to Smells like Teen Spirit.”
“Chris blogged about DiSo.”
“Chris added Steve as a friend.”

Into this:


"At 5:10pm, Chris listened to Smells like Teen Spirit.”
“At At 5:15pm, Chris blogged about DiSo.”
“At 6:30pm, Chris added Steve as a friend.


In the case of music, this might help considerably with the resolution problem that’s been discussed. I could pull out the bits of information from the activity stream that are relevant/interesting to me and do more with them like resolve them to my own catalog (in the case of music). Same could be true with books, movies, videos and other types of media. The aggregate of all of this could be used to build a users personal profile (with their permission of course) or even provide recommendations (note: MyStrands is in the recommendation business). If we could get sites starting to support this kind of format, we could really start to see some extra value applications emerge for developers and users alike.

Where this gets really interesting is when we talk about this within the context of OpenID. When you think about the fact that your OpenID is just an end-point that you have proved ownership of (”I am scott.kveton.com” for example) you can think about landing these annotated activity streams there. Sites and other users will know its “you”. Looking further, with service discovery and things like OAuth, you could feasibly provide private feeds to specific friends or group of people. The possibilities are endless and all of them take advantage of the fact that we’re using very simple building blocks (such as RSS/Atom, microformats, OpenID) to do the heavy lifting. These are the tools that will become the basis for the distributed social network that will be a reality soon enough. Social networks are not a destination; they are simply a feature of every site.

Chris and the DiSo community are starting to discuss these annotated activity streams