Ralph - I can't tell if this a general philosophical suggestion or is in
response to my message of an hour or so ago (which I haven't seen posted yet
so I don't know if you've seen it). Are you responding to the assertion
that "there must be one response schema singled out for mandatory support".
Are you suggesting that there be no schema singled out for mandatory
----- Original Message -----
From: "LeVan,Ralph" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Thursday, December 06, 2007 10:43 AM
Subject: Just Say NO to Anything Mandatory
Here's what I like about OpenSearch: just about any site that supports
searching via HTTP Get can declare itself compliant. I want to be able
to say the same thing about SRU.
It's true that some OpenSearch sites don't deliver much functionality;
they use local query grammars and return peculiar records if not just
plain HTML. But, there is the capability of low-level interoperability.
We know how to get a search to them. We don't know what that search
will do and we don't know how to interpret the results, but we did do
the search. For human users, doing the search is sometimes enough.
Based on that minimal interoperability, they have been able to
demonstrate federated searching capabilities that we can only dream
The level of interoperability is decided by business needs.
In the standards world, specific business needs are encapsulated in
profiles, not standards. For profiles to work well, they need to have
considerable latitude in use of the base standards. Mandatory features
tie the hands of profile developers. We saw this while trying to help
the NISO Metasearch Committee figure out how to use SRU. We ended up
having to say that we were a non-compliant implementation of SRU because
NISO couldn't accept all the SRU mandatories.
So, back to OpenSearch. The only mandatory they have is that you have a
description record. If you don't have that, then you aren't doing
OpenSearch. I think we need to adopt the same philosophy: Explain
records are the only mandatory in SRU. Everything else is subject to
profiling; response formats, query languages, record schemas.
Now, this does mean that there will be sites claiming to be SRU
compliant with whom you can't do anything particularly useful. But all
that really means is that either you don't have a business need to
search them or they don't have a business need to be useful to you. If
either of those does become true, then the minimal SRU implementation
that they have will be better than what they had before they claimed SRU