[Chris Lonvick]
> However, the are exceptions to this behaviour. Potentially, all
> programs with i18n (and/or l12n) support can log messages with a
> different code set. They are many.
Exactly. There are too many to try to standardize on one of them.
One size willl not fit all. This is why i believe the protocol should
be 8-bit clean (ie, accept 8-bit characters), but only standardsize
the first 128 characters.
> If we choose to support different character set we should also
> consider to adopt Unicode encoding. Use simply a ISO 8859-X based 8
> bit character set can turn out to be restrictive.
I do not suggest supporting (or even try to specify) a specific
character set. I suggest allowing almost any character set (most of
them actually have ASCII as the lower 128 codes anyway), and
recommending one of them, ISO 8859/1.
If people want to use Unicode, this would be possible as well, as long
as an 8 bit encoding is used. It is like the date part of the
message; allow almost anything, recommend a sensible default.
If only part of the character set is allowed in the messages, I
suggest specifying an encoding for the rest of them. ie \xxx, &XX; or
some other encoding to make sure all characters _can_ be represented
in the messages. Of course, this is not how syslog currently works,
and should parhaps only be a recommondation -- or left for the next
version. :-)
--
##> Petter Reinholdtsen <## | [EMAIL PROTECTED]
O- <SCRIPT Language="Javascript">window.close()</SCRIPT>
http://www.hungry.com/~pere/ | Go Mozilla, go! Go!
