oh, i don't know did it is a bug or not.

the problem is
my postgresql compile with enable multi bytes = unicode and my default
language is BIG5 encoding, in previous version of postgresql 7.1
i can use unicode encoding to do my back-end encoding and client
encoding set to BIG5, with this setting i can insert a chinese word code
(0xf9d6) in to db (ps: it is not standard of BIG5, it is MS950), another
word (0xea4d) work.
but now
in 7.2 when i insert that word into db, i got a "local_to_utf: could not convert (0xf9d6) BIG5 to UTF-8. Ignored"
it seems very bad for me.
but i have try another back-end encoding, the mule_internal and client
big5 is work with that word.
i remember that word is ok for oracle 8i with unicode encoding.
so that is a standard of unicode,why in 7.2 is fail?

Hard to belive. Neither 7.1 and 7.2 has a mapping from 0xf9d6(BIG5) to
0xea4d(UTF-8). Take a look at
src/backend/utils/mb/Unicode/big5_to_utf8.map.

BTW, if you really want to the conversion between 0xf9d6 and 0xea4d,
you could add it to big5_to_utf8.map and utf8_to_big5.map.
Hope this helps,
--
Tatsuo Ishii