Suppose we define a masking character meaning
"mask zero or more words" (as opposed to '*' which
means "mask zero or more characters")?

Wouldn't that solve most of the index problem?
It would take care of the "first word vs. first
character" problem, and also eliminate the need
for the "Word" indexes, meaning we would need only
one index for each field.

If this seems to be a good idea, someone suggest a
character. I can add this to the 105 proposal
(which is still out for comment, despite the fact
that the date has passed).

--Ray