> > i.e. the index chosen by omission is consistently interpreted
> > independently of the term, but by serverChoice may vary
> depending on
> > the term.
> Nope. _By definiton_, cql.serverChoice does the same thing
> as you get by not specifying an index at all. If
> cql.serverChoice is allowed to choose which index to use
> based on the term it's associated with, as you exemplify
> above (and I think it should be allowed to) then query-terms
> with no index explicitly specified may do likewise. Which
> implies that /explain/configInfo/default[@type='index'] is at
> best misleading.
So if (for a particular implementation) I want ommission to default to
an index as specified in the ZeeRex record, then I can, but
cql.serverChoice must also default to that same index? If I also want to
the clever term dependency stuff in conjunction with omission defaulting
to the index specified in ZeeRex, then I can't use cql.serverChoice but
must use another index (e.g. matthew.serverChoice)?
This distinction isn't just academic - I might want a server which
provides a clever term sensitive any, but also allows omission to be
used as a mechanism for compressing the query (e.g. to avoid long URLs)
without changing the meaning of the query (i.e. allow a client on
inspection of the ZeeRex to know it can safely omit "cql.title" etc.
from the query)