On Tue, 4 Jun 2002, Russell Butek wrote:
> I think we should let the system do it. Always. If I've coded "\n" in the
> WSDL2java code, feel free to slap me around a bit.
The spec tells you to make sure that your implementation send out the
wire values (in bytes) of
GET / HTTP/1.0<cr><lf>
i.e. in Hex
{ x47,x45,x54,x20,x2f,x20,x48,x54,x54,
x50,x2f,x31,x2e,x30,x0d,x0a };
so if you are sure that your
String j = "GET / HTTP/1.0\r\n"
when written out on the network yields those bytes being send out in
exactly that order; *regardless* of the LOCALE the user may have set or
other properties of the compilers(*), JVM, etc - then you are fine.
Which you are in most java(*) environments.
Otherwise you'll need to do something more fundamental when you convert
down to wire format.
Dw
* This is not as trivial if it sounds; some charset's will
substitute the lowercase 'l' for a '1' when mapping from
that to the iso or ascii charset. Likewise some charset
simply lack certain characters; say the H or the K as they
are not used in languages - and you may get country or
language specific substitutions such as to a 'G' or a
'Kappa' symbol.
One particular fun area is for example when quering (oracle) databases
from a locale different than the locale programmed/entered in. And even
though all is UTF-8/unicode; a 'e'-trema may be receid back as a syngle
code point -or- as a 'e'-'backspace'-'trema'. Or a cyrillic 'c'
may end up being another code point; etc. I.e. they all look as glyphs
the same to the user - i.e. their shape - but their codepoints or
sequence of bytes interally may differ from what the programmer expected.
