Mike Taylor wrote:

> My immediate feeling was that a prox-based solution would just scare
> everyone off because, well, prox is scary.  But maybe people feel the
> same way about relation-modifiers, too.

Proximity search is not scary but rarely implemented, documented and 
used. There are some workarounds

http://www.staggernation.com/cgi-bin/gaps.cgi
http://www.researchbuzz.org/2004/10/ynaps_yahoo_nonapi_proximity_s.shtml

but you really have to know it before to find out how to use proximity 
search. I wonder why OPACs and search engines don't show searching tips 
like "hey, you have a lot of results! Try proximity search!" more often.

> Maybe the pragmatic approach is to introduce a new boolean, "with",
> which is explicitly defined to nothing more or less than an
> abbreviation of prox/unit=element/distance=0.  Then server
> implementors can tackle "with" free of Proximity Fear.

Nobody will type in "prox/unit=element/distance=0" by hand. The common 
expression for proximity is "near" so please don't introduce something 
new. For the end-user you better rewrite CQL queries anyway
(for instance 'cat 2 words before hat' becomes '"cat" 
prox/distance>2/unit=word/ordered "hat"') so 
"prox/unit=element/distance=0" should be fine for implementors.

Greetings,
Jakob