>>I want the CQL syntax to be a naively simple as possible, even to the
>>of losing some functionality.
I fully agree with the above. I like the original CQL definiton (including
the quoted search terms) because I think it is important that CQL queries
can be handled by most servers (and users).
From experience I know that parsing the quoted query terms is impossible
without full knowledge of the underlying index definition. I feel on the CQL
level we should only specify which wild cards are allowed within terms and
leave the actual interpretation of search terms to the server. If the user
wants certain unambiguous behaviour he can rephrase the query.
The mapping of CQL terms to Z39.50 queries is an implementers problem! As
part of for example the bath profile for certain key types constraints on
server behaviour are possible but there are limits. In the case of UTF-8
indexes simple things as "what is a word" are hard to resolve and truly
outside the scope of CQL. All the same when the user types "Tasmanian tiger"
he can reasonably expect that the server tries to find records containing
the phrase "Tasmanian tiger".