On Mon, 2003-10-06 at 10:24, Albert Mietus wrote:
> Rainer Gerhards writes:
>
> > TAG = full-stat-id [full-dyn-id] (':' / SP)
> > full-stat-id = [path] progname
> > full-dyn-id = '[' proc-id [thread-part] ']'
> > path = path-part 1*(path-sep [path])
> > path-part = 1*VISUAL
> > path-sep = '/' / '\'
> > progname = 1*VISUAL
> > proc-id = 1*ALFANUM ; recommended: number
> > thread-part = thread-sep thread-id
> > thread-sep = VISUAL ; recommended: ",", or ':', or '.'
> > thread-id = 1*ALFANUM ; recommended: number
> > VISUAL = ([a-zA-Z0-9...], excusing '['
> > SP = %d32
>
> Some remark (aside it look nice)
>
> * See my other mail about the termniting colon.
> * The terminating SP shouldn't be part of the tag. I think, it is a
> separator between fiels/parts (like in other parts of -sign.
I agree. If you look at my original comments, I said that I just
included it as above because there is no way you can do this in ABNF
without specifying more than just the tag. I will try to update my
(partly not published yet) syslog ABNF to include your sample and then
post this.
> * I agree on you other mail: proc/'/thread can be main/detail
> * A path may start with a path-part (/like/this)
>
> > By updating it, some more points were raised (it is much like
> > implementing;)):
>
> > - I made full-dyn-id optional
> Yes
>
> > - VISUAL must allow non US-ASCII characters
> It shoud allow allo characters that are allowable in "syslog" :-)
In RFC3164 (as well as e.g. sysklogd real world package), they can be
non US-ASCII. So this would just save from further limitiation not
currently existiing.
>
> > - do we really need to describe the path?
> Yes, Bit limmited. Only to describe how to find PROGNAME. Which is
> frequently used in relays (all syslogd use it. The use PROGNAME, not
> TAG!
OK, I think a language issue. My point was not if we need a path. It is
helpful. My point was if we really need to describe it in ABNF:
> > path = path-part 1*(path-sep [path])
> > path-part = 1*VISUAL
> > path-sep = '/' / '\'
>
wouldn't it be sufficient to say
path = 1*VISUAL
thus not thinking about path being separated by slashes?
> > - why not fix thread-sep to ","?
>
> Allowing "many" characters, make it possible to use "native" notation
> for several platforms. (I don't use thread frequently, so I made a
> guess which chars to uses).
>
> It is like allowing both '/' and '\' for directories. (Allowing ':'
> for the Mac is complex, as the colon becomes used to often.
Ah, yes... I now agree.
>
>
> > - MUST proc-id/thread-id by ALFANUM - would a SHOULD be sufficient (what
> > do we gain from a MUST)?
>
> Probally, I'm not an expert on MUST and SHOULD notation, I must ( or
> should:-) say.
If I get it right, MUST is it really MUST - always, ever, no way around
it. SHOULD is a recommendation, but it leaves room to treat do it
differently if the implementation sees a need for it.
>
> > - The ABNF is still not fully correct in the terminating SP of the
> > message... This can not be done only by specifying the tag. But I think
> > it is sufficiently to convey the idea.
That was the (somewhat hidden;)) point about SP not being correctly
represented ;)
>
> See above,
>
> Thansk for the help!
Thanks for all your help!
Rainer
