I do not want to see empty query parameters signaling searchingRetrieve!

I see no problem with a rule that says either the presence of a query parameter or a queryType parameter signals that we have a query. The queryType will then have some associated parameter(s) that carry the query. That parameter might even be the query parameter.

Ralph

From: Hammond, Tony [mailto:t.ha...@nature.com] Sent: Thursday, December 16, 2010 6:08 AM To: Hammond,Tony; Ray Denenberg, Library of Congress; Matthew Dovey; LeVan,Ralph; OASIS SWS TC Subject: RE: [search-ws] queryn: Some further comments

Hi:

Some further comments.

1. On reflection I guess that "queryn" could be dispensed with as long as an empty "query" parameter were present in the query to a) signal the searchRetrieve operation, and b) to receive the assembled query.

2. I had earlier proposed briefer forms of the parameters which would make the querystring more manageable.

Instead of

"q1.idx", "q1.rel", "q1.trm", "q1.bln"

I proposed

"qi1", "qr1", "qt1", "qb1"

which are cleaner (half the length) and slighly easier to manipulate (only prefix and index to manage instead of prefix, index, and suffix).

3. Since the query parameters are built with integers they are arbitrarily extensible (do not require a fixed length form) and sort order is completely in hands of the form generator. Doesn't matter where the parameters appear on the form - just their call order. And then they can be simply iterated over.

4. My preference is still for an explicit count (queryn) as I think this is easier than trying to compute the number of clauses. And the count is already known to the form generator since it is adding an index to the clause parameters. Would almost seem to be a no-brainer.

5. Because of the boolean coupling of search clauses we require a boolean associated with the search clause which for convenience we take to be trailing - i.e. operates between this clause and any subsequent clause. The final boolean in the series will always be omitted.

Tony