RE: Betr.: RE: Qualifier Combinations in CQL (Was: Bib-2 and the DC-Lib)

I think that there are two quite opposite approaches in searching: 

1) doing a very precise query and knowing exactly what will be 
returned (more for the professional? user that knows xactly  that he 
wants to search for  "creator is not author and has an x in the third 
postion of the second field of  DC.subject but not Bib1.subject)

and 

2) the more associative approach where users SEARCH and do not 
know exactly what they want or how to search. I do not want users 
to be punished for searching in an index that does not exist.
 
Searching for (author:"Levan" or notExistingIndex:"van Veen") can 
result in:

1) only author:Levan and a field in the response indicating that 
notExistingIndex does not  exist
2) only author:Levan
3) Error message

I want to serve users with option 1) or 2) and you (Ralph) want to 
serve the user  with option 3).  

You do not want a client to keep trying different indexes until he 
gets a non-error response. Clients for the "dummies" should try both 
"fuzzy" and "any" because "fuzzy" returns an error if "fuzzy" is not 
supported even if there is an exact match.

I do not see a solution right now for the two opposite solutions, but I 
am convinced that we should try to get both approaches into one 
standard to offer users the choice for both approaches for each 
individual query.

Theo


On 25 Apr 02, at 10:35, LeVan,Ralph wrote:

> > -----Original Message-----
> > From: Theo van Veen [mailto:Theo.vanVeen@kb.nl]
> > Sent: Thursday, April 25, 2002 10:12 AM
> > 
> > Does this mean that you are not allowed to send a search 
> > before asking explain what is supported? And when you do for 
> > example (author:"Levan" or notExistingIndex:"van Veen") does 
> > the server give an error message or does it return the hits 
> > for author:"Levan" , or does it both? 
> 
> I'm guessing that a static copy of the Explain record is good enough,
> at least until something that used to work no longer works.
> 
> notExistingIndex:"van Veen" definitely gets you a failed search.
> 
> Ralph

Received on Friday, 26 April 2002 21:59:32 UTC