RE: JW10: QSD usage

At 10:18 AM 8/10/99 -0700, Niket Patwardhan wrote:
>Two different search engines are going to return different documents
>(that's reality). Even the same search engine will return different
>documents if set up to interpret the document differently. (For example, if
>they have different thesaurii) 

I agree, this is my experience as well (and it is echoed by others on the
design team)
Do you see anything in the protocol that is in defiance of this?

>But it would be nice to ask "Does this field
>exist, and if so, pay attention to this part of the whole query". So it
>could be acceptable.

Yes, DASL defines an operator that returns TRUE iff the field is defined.
So if for example you want a query that finds all documents by Reed
Richards that are longer than 10Kbytes,  but if the server does not even
know the length, you would rather have all documents than none, you can say
this.

>But I think what you are getting at is that we need to identify the
>conditions and a subset of the full query language which an expert knows
>would get exactly the same answers from all servers.

Hmm, that seems hard.  Perhaps if one limited one's query to boolean only,
but then that would exclude all the good stuff that search engines like e.g
WAIS and Excite provide.  

QSD aims are more modest.  If I am using a DASL client with a good GUI, I
should be able to visit a new DASL server, about which I know nothing, and
if it has a good implementation of QSD, I should be able to generate a good
query using any extension it supports.  Example, I visit a book catalog and
discover it not only has author and title but also has a field called
"favorably reviewed by".  Then I can make a query for books by Richards
that are favorably reviewed by *Patwardhan*.

Does this seem a reasonable goal to you?

Jim

Received on Monday, 16 August 1999 20:14:15 UTC