On Mon, May 30, 2005 at 07:35:22PM -0400, Robert Sayre wrote:

> >I get the feeling that OpenSearch + Atom could be real useful.
> 
> There is substantial overlap with Atom Protocol paging, especially
> existing gregorio-07 implementations. I quite like their approach to URI
> construction and their description document, which allows servers to 
> place the query parameters anywhere in their URI space.
> 
> http://opensearch.a9.com/spec/opensearchdescription/1.0/

I seem to remember when I looked at it that it rang bells with me by
looking like part of WSDL. At which point I wondered why they didn't
use WSDL. That's more of an aside, though.

A couple of us from Xapian [1] looked at OpenSearch, and implemented
it for our example search engine. It took about half an hour (mostly
because we use a templating language for query results), but ten to
fifteen minutes of that was spent figuring out what OpenSearch
actually meant, and we had to make a number of assumptions in doing
so.

Were someone to produce a document describing Atom+OpenSearch, I'd
expect it to have the rigour of Atom. OpenSearch is written in the
RSS2 style, which isn't really appropriate for a highly interoperable
format, which is clearly what's desired.

We were also a little concerned that the OpenSearch model was very
simplistic - there were many possible use cases we could think of that
weren't catered for, and indeed we could only think of one that was
really supported, being the one that A9 puts it to (which is a slim
superset of feeding OpenSearch into a desktop aggregator). We did
start to put together a proposal for an Atom+Search extension, and
completed the conceptual work but then got distracted by other
things. This is kind of orthogonal to the OpenSearch issue, but if
people are interested in discussing a richer search extension we can
try to clear some time to pull it into shape.

[1] <http://xapian.org/>

James

-- 
/--------------------------------------------------------------------------\
  James Aylett                                                  xapian.org
  [EMAIL PROTECTED]                               uncertaintydivision.org

Reply via email to