Hi Anton,

Thanks for your comments! I am answering a bit out sequence...

> RFC2277 "IETF Policy on Character Sets and Languages" could be of use
> for syslog-international draft:
> ftp://ftp.isi.edu/in-notes/rfc2277.txt
>

I went through it while drafting...


> Of note:
>
>  - It clearly favors UTF-8.  Does not mention UTF-7.

The issue here is that UTF-8 is 8 bit. Unfortunately, all other syslog
RFCs/Ids specify the message to be 7 bit and US-ASCII only. I already
sent a note to the WG list if that should be changed in the other RFCs -
responses are still outstanding (I guess vacation season is hitting...).

The issue here is even when it is changed in the other RFCs, existing
syslogds may be broken by 8 bit characters. So this may not be something
that could be done easily. On the other hand, at least my experience
shows that most of them handle 8 bit, so it may be worth re-thinking.

In this position, I selected UTF-7 because it can encode within the
US-ASCII constraint.

I would really appreciate any comments on the other RFCs, too -
especially 3195 and syslog-sign. If they had 8 bit chars,
syslog-international could indeed plainly specify UTF-8, which would
resolve many issues... But again, we must ensure we do not break
existing things! One goal behind syslog-international is to make it
compatible with existing syslogds. As we all know, changes move in
gradually and I expect that "older" syslogds will be around for *quite*
a while...

>  - It seems to argue that language must be identified.

This is why it is in...

Rainer


Reply via email to