How open is OpenSearch?

Last week at the O'Reilly Emerging Technology Conference, Jeff Bezos announced OpenSearch, an API that enables third parties to inject their own live search results into Amazon's I didn't attend ETech this year, but that cloud had a silver lining: I was able to dive right in and do an OpenSearch implementation.

As fate would have it, I'd just finished a quick hack to reorganize the output of InfoWorld's Ultraseek search engine. I named my little project InfoWorld power search because it delivers a dense page of titles, categorized by story type. When I heard about OpenSearch, I wondered how hard it would be to integrate my new view of InfoWorld search as a "column" in A9. As I soon learned, it's almost trivial.


OpenSearch is interesting in lots of ways, but here I want to focus on its use of RSS. A9 doesn't subscribe to my search-results feed in the way Bloglines or FeedDemon or NetNewsWire would. It doesn't poll for changes. Instead it sends a request to my site when an A9 user with an active InfoWorld column performs a search. The response packet I send back just happens to be formatted as RSS 2.0, but from A9's perspective, it could be any XML format.

Why RSS 2.0, then? Because it creates network effects that go way beyond the point-to-point relationships between A9 and its search partners. The work I did to export RSS 2.0 search results served double duty. It accomplished the integration with A9, but it also dramatically expanded InfoWorld's RSS surface area. Now, for the first time, you can subscribe to any InfoWorld search in a feed reader. Want to be notified when the next review of a VoIP product shows up at Run the query, and subscribe to its results.

Most people nowadays use RSS for person-to-person communication. You know the pattern: When a publisher posts a blog item, subscribers are alerted. A growing number of folks are also using RSS for process-to-person communication. Subscribing to searches is the best example of this pattern.

A9's use of RSS for process-to-process communication represents a third pattern. We'll be seeing a lot more of it. Not because RSS enables process integration in special ways -- it doesn't -- but rather because RSS helps us blur the boundaries between human network and process networks.

To be honest, I wasn't even planning to enable RSS subscription to InfoWorld search. It just came for free. When that happens, it's a sign that things are deeply right. [Full story at]

In response to this column, Raymond Yee asked the obvious and important question: Are there any other public registries of OpenSearch implementations? I'm not aware of any others, but so far as I can see there's nothing preventing that.

As a matter of fact, I'm thinking of doing one myself to help advance the state of structured search and structured blogging. My thesis has always been that these activities depend on one another. Without structured search there's no reason to produce microcontent, and without microcontent there's nothing interesting about structured search.

Here on my own blog I've been working both ends of the street for a couple of years. My local structured search feeds off the enriched XHTML/CSS dialect that I use when I write my own blog entries. My extended structured search widens the scope to the hundreds of blogs I read -- though in that case there are few if any other microcontent nuggets to find, since nobody I read seems to be producing them.

I have, however, been talking with a few folks who are in a position to build similar search implementations. If that happens I'd like to federate them into a network that will incent the creation of microcontent.

As I mentioned last week, I don't think formats are the gating factor. Responding to last week's item about PubSub's structured blogging initiative, several folks pointed to prior art. Indeed, there's plenty of it, this isn't a new idea. Tantek Çelik, for example, has been thinking along these lines, and Adam Rifkin posted a nice analysis of Tantek's approach. Plenty of people have been down this path. For me, the CSS-oriented approach I'm taking has roots that go farther back than my 1999 book.

Lots of folks have chewed on this problem. But the fact remains that we have yet to kick off a virtuous cycle that rewards microcontent creators, directly and immediately. That will require:

  1. A way of producing microcontent that's easy and natural for average folks.
  2. A way of harvesting microcontent that's easy and natural for average folks.
It seems likely that the first requirement will be met by plugins to blog-writing software, and it was PubSub's plugin strategy, not its microformat, that I found most interesting.

It's certain that the second requirement will be met by aggregators and search engines that are microformat-aware. So, back to Raymond Yee's question about OpenSearch. If I want to join a federation of microformat-aware aggregators and search engines, maybe even host its registry, does anyone know a reason why I can't use OpenSearch -- independently of A9 and -- for these purposes?

Former URL: