Is the assumption in 2.1.2, that the same query submitted
multiple times will return the same results, a safe one? I don’t
think so. I’d probably say that the result set model assumes that
the server throws away the results after the response and that the client must
be prepared for different results if it submits the same query again.
2.1.3: Drop the gratuitous comment about an item sometimes
being referred to as a “result”. I’ve never seen that
and don’t like it.
2.1.4: “If the
server is able to interpret but not process a request it can send back the
OpenSearch Description Document that explains how to correctly construct a
request.”. Is this codified behavior or have we invented
it? Unless you can point at that behavior in one of their documents, I’d
prefer to drop it.
2.1.5: Item 6: I think this note misses the point of
Auto-Discovery. The specific example given is probably the least
interesting. More interesting are the ways that you might find an
OpenSearch description document in other contexts, such as in an HTML or RSS
response.
2.3: I think it is worth saying that the elements listed are
associated with a specific namespace and list that namespace. You should
also say that those elements assume an XML response, but that an HTML response
with none of those elements present is perfectly acceptable. Behavior for
other mime-types is undefined (and may even be illegal?).
3: I believe that mime type should be in all lowercase.
3.1: This document is supposed to be a binding to
OpenSearch. I don’t think we have any business defining their
description document.
Ralph
From: Ray Denenberg,
Library of Congress [mailto:rden@loc.gov]
Sent: Friday, October 09, 2009 1:57 PM
To: OASIS SWS TC
Subject: Re: [search-ws] opensearch
-----
Original Message -----
Sent: Thursday, October
08, 2009 5:56 PM
Subject: [search-ws]
opensearch
The
most recent version of the OpenSearch binding (February 21) is:
Tony,
are you working on that?