Multiple URL Construction Models

What we will produce can be simply described as instructions for how to construct URLs for talking to search systems and clues about how to interpret the response. I’d like to talk about that first part.

I’m concerned that we tend to think in terms of SRU and think of URL construction as tacking parameters onto the end of some opaque URL base. That is not the only way that searching is done. Often, search terms are embedded in URLs. Try http://www.worldcat.org/identities/lccn-n98-1479. That URL returns an SRU searchRetrieveResponse. I’m going to want to be able to describe that server with our new description record (I’m not sure we’re going to continue calling it an Explain record.)

OpenSearch uses a URL template and describes how to interpret placeholders to be replaced in the template. That is probably a good model for cases like mine. It has weaknesses as we scale up in complexity. What happens when I have a concept in my URL for which the client has no appropriate value? Does the template then need to have a placeholder for every concept it understands?

Another model is to build up a base URL with parameter=value pairs. Unfortunately, the REST community doesn’t like that behavior.

I suspect that we’ll need to accommodate both models going forward.

Ralph

search-ws message