Yesterday I talked a bit about converting all my documents into HTML and then running a search engine over that. I also lamented that it would be easier if all the search engines were part of the Well-Formed Web. What I mean by that is using multiple search engines and combining the results would be much easier if they all were able to format their results in XML.

I was still kicking that idea around today when I started thinking about what would be the best format to use for search results. For example: here is a search over the swish-e discussion list, the output being in HTML of course. But look closely at that output and tell me what you see. I see:

  • title
  • link
  • description
  • message date

And because this is a search over a discussion archive I also see:

  • poster's name
  • poster's e-mail

A similar search on Google yields a subset of the same fields but adds in

  • category

Maybe it's just a "when you have a hammer, everything looks like a nail" situation, but to me Atom looks like the perfect format for returning search results.

Note that search results are also an excellent place for some of the features of the AtomAPI that have been talked about recently. For example, most search results don't return all the results, but only the first N. To get the next N results the Atom feed format could contain a "next" link that points to the next N search results.

<link rel="next" type="application/x.atom-xml" title="Next 20 Results" href="..."/>