>I'm afraid it doesn't seem to be SRU if you return HTML.  When you
>return a stylesheet, you still return XML and have the browser render
>the XML to HTML.

>I'm afraid that you're closer to OpenSearch than you are to SRU, but
>OpenSearch doesn't define a standard query grammar.
>Rob Sanderson has had discussions with the OpenSearch folks about
>support for CQL.  He might have a suggestion for you.

It's possible to go all the way apart from the very last step into
SRU compliance with OpenSearch.  In fact you could think of SRU as a
very specific profile of OpenSearch.

My suggestion as to how to do it:

Have one end point that handles both SRU and HTML responses, but give a
parameter to determine which is required.

Name all of the parameters according to the SRU naming scheme, such that
your 'HTML please' parameter would be something like &x-html=1

Then create an OpenSearch Description document which gives both mime
types for the responses.  Which of course brings up the question of what
the mime type for the SRU response is... My suggestion is:


<!-- XHTML via OpenSearch -->
<Url type="application/xhtml" xmlns:cql=""

<!-- Full SRU compliance -->
<Url type="application/sru+xml"

Then the people who want html to import directly into a web page can
have it via opensearch, but they're only a couple backspace clicks away
from getting the full SRU response.

DeWitt and I will be publishing an 'official' How-To for OpenSearch/SRU
shortly (this week is my last week I have to give lectures this

Does that help?