> No, I'm suggesting that it shouldn't be let loose on Big5 data when it
> evidently hasn't a clue about that encoding. The byte in question
> *is not* a backslash, it's not even an independent character; and so
> changing it on the assumption that it is logically a backslash simply
> breaks the data.
Would you please enlighten me the behavior of the backend - why
SET CLIENT_ENCODING TO Big5;
INSERT INTO y VALUES ('A\134B');
stores
A\B
while
INSERT INTO y VALUES ('y\134na');
--"y\" and "na" are two Big5 characters.
stores
y\134na
instead of
y\na
> Your quickest route to a solution may be to avoid Big5 in favor of
> an encoding that is ASCII-safe, such as UTF8. You can feed that through
> code that only understands ASCII with much less risk than an encoding
> where second and later bytes might look like ASCII.
Are you suggesting me to implement the middleware that will translate
Big5 input to UTF8 and then escape the latter before sending it to
PostgreSQL?
SET CLIENT_ENCODING TO UTF8;
[BIG5 string from user] --> [middleware] --> [UTF8] --> [escaped UTF8]
--> PostgreSQL (initdb with -E UNICODE)
Best regards,
CN
--
http://www.fastmail.fm - Does exactly what it says on the tin