In the limited spare time that i seem to have these days, one of the things i've been looking at is OpenSearch, "... a collection of simple formats for the sharing of search results".
OpenSearch is slowing gaining lots of traction and each of the major browsers now allow you to add an OpenSearch to your toolbar automatically (although the User Experience certainly needs work!).
In my current consultancy work for the NHS I was asked to extend the search API I wrote some months back which works against FAST, and enable any third party site to ended some script in their site and make Ajax style queries. Last week I wrote this functionality (which returns custom Xml and Json) and sure enough it works great.
However, I like open stuff because it makes it easy for others to pick up and run and so after I added OpenSearch support via the usual description file and tested this with Internet Explorer et al and it worked, i decided to look at whether i could use OpenSearch as the actual syndication format.
Now, if this all ran on the same domain there would be no problem in pulling back Atom or RSS and formatting it for display. But in a cross domain architecture you can't do this due to Same Origin Policy. So, using JSON-P we can enable OpenSearch queries over the wire.
To do that however, we need to have an OpenSearch format for JSON - in my case i simply took the Atom format as a starting point and mapped much of that and added in the OpenSearch elements and attributes.
So now you can take a few lines of code, add it to any page in the world and make an OpenSearch query using the template Uri and get an OpenSearch response which can be parsed and dislayed.
Early days but if it's useful to no-one else on the planet, it is useful to me :)
If interested, please try it out here. It queries over 100 million records so you will get results no matter what you type in, but the formats and data mappings are still being worked out.
PS. Would be *very* nice if we could query the OpenSearch discovery endpoint and ask for JSON to be returned. That way we could dynamically build all of the front end stuff that i can't do with the existing reponse (it returns an OpenSearch specific content type which is not trusted by browsers).