>Hmmm. It seems to be a combination of two problems:
>
>1) not knowing which type should be preferred (or how to compare varchar
>and bpchar; should I strip the blank padding from a bpchar? Probably
>so...).
>
>2) swallowing the type coersions since bpchar, varchar, and text are
>considered to be binary compatible. I may have seen another instance of
>this being a problem, and perhaps should figure out how to propagate the
>new coerced type into the query rather than dropping it.
>
>Will look at this. Question: how *should* we compare bpchar and varchar?
>It may be that we should have some explicit comparison or coersion
>routines to make things work smoothly.
Not sure. I will check some SQL books at home.
--
Tatsuo Ishii