[Python-Dev] should we keep the \xnnnn escape in unicode strings?

[Guido van Rossum]

> I'm with Tim here: screw the retarded compatibility. Make \x make sense in
> normal strings and unicode strings both, but in a way that breaks as little
> as possible: if it always results in one byte in 8bit strings, it should do
> the same, IMHO, in unicode strings. I'm also for making \x take at most four
> bytes, like the documentation states, instead of hassling with a perl-like
> construct such as C has. I would be very, very suprised if anyone was using
> \x29E8F28DA727 and expects it to work like \x27. 

Haven't read Tim's post, but I agree to screw compatibility here, and
I see two reasonable ways out:

(1) \x takes up to 2 hex chars in 8-bit strings and up to 4 hex chars
in Unicode.

(2) \x takes up to 4 hex chars in all strings.

The presence of -U and ease of JPython compatibility both suggest that
(2) is best.

--Guido van Rossum (home page: http://dinsdale.python.org/~guido/)