> Date: Tue, 29 Nov 2005 11:46:16 -0500
> From: "LeVan,Ralph" <[log in to unmask]>
>
>> If I allow for the specification of a value I need to know, don't
>> I, what to do when a value is given.
>>
>> What does a distance=10 mean for some arbitrary element? How do we
>> view order or direction?
>
> I've built systems with proximity across elements. Imagine this case:
> <rec>
> <author>C. J. Cherryh</author>
> <title>Brother at Arms</title>
> <author>Terry Pratchett</author>
> <title>Small Gods</title>
> </rec>
>
> I'm looking to see if Cherryh wrote "Brother at Arms". The search
> is: Author=cherryh prox/unit=element/distance=1 title="brother at
> arms"
Right. But more importantly there is a principle at work here, which
is that we should _never_ prohibit something just on the basis that we
can't immediately see a use for it. That's the kind of thinking that
almost led us a while back to outlaw zero-length query terms in CQL --
a decision which, had we taken it, would have made it impossible to
scan from the start of an index, among many other negative outcomes.
It's often been said that one of the hallmarks of great software
design is that it's used in ways the designers never envisaged. It am
prepared to bet that in years to come, people find all sorts of uses
of CQL proximity that haven't yet occurred to any of us on this list.
(And as Mark H. has demonstrated, there _have been_ historical uses of
it that no-one on this list _remembered_! :-)
> Date: Tue, 29 Nov 2005 21:34:19 +0100
> From: "Edward C. Zimmermann" <[log in to unmask]>
>
>> We don't always get to define how the records we get are structured.
>
> But you are, de-facto, also demanding that the user also know and
> understand how the records you get are structured to be able to
> search them.
We're not _demanding_ anything. All we're doing is providing
facilities so that _if_ the user knows the structure, he can address
it the queries. (And the example we started with shows that, in some
cases, the user absolutely does know the structure and wants to
address it.)
> Date: Wed, 30 Nov 2005 07:19:13 -0500
> From: "LeVan,Ralph" <[log in to unmask]>
>
> > marc.500$a any "baz" prox/unit=parent/distance=0 marc.500:1 = 1
>
> That's pretty much what I would expect it to look like.
I am also happy with that.
... though I would still like an alias "with".
_/|_ ___________________________________________________________________
/o ) \/ Mike Taylor <[log in to unmask]> http://www.miketaylor.org.uk
)_v__/\ "We should not hold rashly an opinion in a scientific matter,
so that we may not come to hate later whatever truth may reveal
to us, out of love for our own error" -- Saint Augustine.
|