Guido van Rossum
Fri, 07 Jul 2000 16:58:56 -0500
> That is not my understanding of the agreement, and I think the whole
> idea of a default encoding is a stupid one. I strongly suggest that
> the behaviour that Ping just explained and that Guido had suggested
> originally is implemented: Mixing unicode and string objects is only
> allowed if the string objects are ASCII. Everything else (including a
> "default" encoding of some kind) is an application issue.
> Otherwise, you'll just go on and ask questions endlessly how things
> should behave, and change the interpreter forwards and backwards.
> The meaning of mixing both kinds of objects is very easy to explain,
> and also allows for efficient implementation - which you won't get if
> you have to invoke a character conversion just to get the hash value
> of a unicode value (!)
I agree with this point. It can be implemented easily by deleting the
locale specific guessing from site.py.
--Guido van Rossum (home page: http://dinsdale.python.org/~guido/)