Right.. we have in the big picture more than just date ranges
and trying to provide a mechanism for clients to "understand" them is
complicated--- and few will ever bother--- if not impossible.... but also
delivers little. For general facet search we need shared sematics for things
anyway to make sense.. If a user at a clothing shop selects "hose"---
a word that means pants in German--- they want the items that are pants and not
fire hoses--- the "Fire Hose" gag from Bill Dana in his guise as Jose
Jimenez comes to
mind.
The anything goes model does not mean that one can't return (or handle) ISO defined date ranges if one wishes.. All we have done is say that what the server provides in a scan or for a facet it must understand and do what it intended when it supplied them. This is, I think, already a step above the supply a term "exactly as its in the index" without a explicit contract to do (or be able to do) something meaningful (from the perspective of the server) with them.
And remember.. Date ranges are not the only business case for this approach.. Sure.. if all you have are full text terns, dates, date ranges, numbers and number ranges one might be able to devise something but once we start getting into the realm of more obscure data types and ranges things get weird.. Do we really want to start to talk about masking etc.. ISBN ranges anyone? Closed paths..
On Mon, 25 Oct 2010 16:13:14 +0200, Edo Plantinga
wrote
> We
*don't* have client-defined range facets, therefore
the developer cannot figure out how to create such a query anyway. Your argument
does not hold true for server-defined facets. To put it another way: there will
be no sending of strings that have not been sent first by the *server*, and
therefore there will be no "url hacking" or "query
hacking".
>
>
Van: LeVan,Ralph [mailto:]
> Verzonden: maandag 25 oktober 2010
16:02
> Aan:
> Onderwerp:
[search-ws-comment] "Last Week" is a bad idea for
<actualValue>
>
>
> I’ve been giving
more thought to our facets conversation
and have decided that I don’t like “Last Week” as a term to
be sent back to
the server. I’m not saying it is illegal or that the standard
won’t
support it. I’m just saying I think it is a bad
idea.
>
> The reason is that it depends on server magic. The
client, or more importantly the developer, won’t learn anything about
how to
construct other range queries if we hide how it is done behind magic
strings. If, instead, we send “20101017 20101023” as the
<actualTerm>, then the developer might be able to figure out how to
create their own query for “Two Weeks
Ago”.
>
> Of course, an <actualTerm> of “20101017 20101023”
would want a <displayTerm> of “Last
Week”.
>
>
Ralph
That's actually evil since that DisplayTerm is volatile while the actualTerm is persistent.. By the time the query is sent off it might not be "Last Week" but "2 weeks past"..
--
Edward C. Zimmermann, NONMONOTONIC LAB
Basis Systeme netzwerk, Munich Ges. des buergerl. Rechts
Office Leo (R&D):
Leopoldstrasse 53-55, D-80802 Munich,
Federal Republic of Germany
http://www.nonmonotonic.net
Umsatz-St-ID:
DE130492967
