|
Well then we need more text to explain all
this. Ralph, would you write a section "Scan Clause" that in particular
explains how the relation works, with specific examples. It should also
explain (or somewhere the document needs to explain) the relationship to
CQL. If the scan clause is the CQL seach clause, then does this mean
that scan is to be used only (or primarily) when the follow-on search uses
CQL? Also, the agreement you cite about how to scan from the
beggining of an index, where is it? It needs to be in the standard, I
suppose in the (new) "Scan Clause" section.
Thanks.
--Ray
---- Original Message -----
Sent: Tuesday, November 10, 2009 10:34
AM
Subject: RE: [search-ws] Scan
proposal
I’m
strongly in favor of keeping all the features as they are, as they are all
being used by me.
Yes,
keep scanClause. The relation is important, just as it is for
searching. A search for title=”dog” and title exact “dog” retrieve
completely different things and a scan for those clauses retrieve different
terms. The first would retrieve a list of single words and the second
would retrieve complete titles.
We
have an agreement for how to scan from the beginning of the index: provide an
empty scanClause. Scan title=”” will get you the beginning of the title
words index.
No
changes are necessary or desirable.
Thanks!
Ralph
From: Ray Denenberg,
Library of Congress [mailto:rden@loc.gov]
Sent: Monday, November
09, 2009 5:06 PM
To: OASIS SWS TC
Subject: [search-ws]
Scan proposal
I'm working a
bit on the Scan document. I don't think we have ever had a serious
discussion about scan within this committee and I would like to raise the
following issue (that comes up from time to time on the SRU list, but I don't
think has been discussed here):
Do we really need
to retain the "scanClause"? Could we instead define separate
parameters for index and start term?
The scanClause was
defined as such in the early scan version in order to re-use the seach clause
definition from CQL (and actually the history goes back further than that -
there was originally a Z39.50/asn.1 analogy that got carried over to SRU) and
I'm not even sure that this "re-use" ever even made sense. But in any
case, don't we want to decouple Scan from CQL? Or do
we?
Another reason: the
Scan clause includes a relation, which I think is completely superfluous.
Nobody has ever explained, without much pain, how any relation other than
'equal' makes sense in a scan clause.
A third reason: we
had talked about (ok, I suppose at sometime we did discuss Scan, then) the
start term being optional, so if omitted would default to the first term in
the index - say that someone wants to scan from the start of the index but
doesn't know what the first term in the index is, it is difficult to formulate
the request when the start term is mandatory.
Could we at least
define both a Scan 1.2 and Scan 2.0, and make this change in
2.0?
|