On Sat, Dec 28, 2002 at 01:41:34AM +1100, Darren Reed wrote:
> In some email I received from Bennett Todd, sie wrote:
> [...]
> > I think the next step we need is to survey existing sylog-over-tcp
> > implementations and see who is terminating their records how, and
> > what sort of interop (if any) is available.
>
> Well since someone has asked, I'll volunteer some information on what
> nsyslogd does over TCP :)  The below is part of a text file from it.
> I'm not aware of anything being interoperable with it.  Log messages
> are limited to somewhere just under 64k and since the message size is
> passed through, there is no record delimeters used or required.
>
> btw, I don't think this is perfect (ms or finer resolution might be
> nice for timestamps, for example).

Thanks Darren. Andrew has already sent me a proposal about his protocol,
would you please resend it to the list?

I think it was also Andrew who summarized the PIX TCP protocol:

# Sends on TCP port 1468 by default.
# Can have multiple messages in a single packet. They don't appear to be
seperated by any character. New message is identified by the <PRI> tag.
# TCP connection is made once, and remains open indefinately.
# If the connection is broken and not able to be re-established, the PIX
will stop forwarding network traffic through it's interfaces.
# PIX message format changes depending on timestamp settings on the PIX.

I think messages should be terminated somehow, probably with a '\n'
character? Anyone with a PIX at hand can confirm this? This seem to be
compatible with my previous syslog-over-TCP description (which described the
way syslog-ng behaves).

-- 
Bazsi
PGP info: KeyID 9AF8D0A9 Fingerprint CD27 CFB0 802C 0944 9CFD 804E C82C 8EB1

Reply via email to