On Sun, Dec 22, 2002 at 11:03:19AM -0800, Darren New wrote:
> Balazs Scheidler wrote:
> > The description should focus on the means of encapsulating one single
> > message into the TCP stream, where the message has the same format as the
> > UDP packet in RFC3164.
>
> [snip]
> Just FYI. The problem *has* been solved. :-)

I know about that. I don't want to enter endless debates again. syslog/tcp
is not competing with syslog-reliable. syslog-reliable will be the future,
syslog/tcp is what current products do. In fact, TCP transport is the most
useful feature of recent syslog implementations, maybe their only reason for
existence.

Simply using syslog/raw and a trivial beep implementation is like using
an SQL server to store your 100 user accounts. Syslog benefits from the
advanced BEEP features, but the RAW profile gives you nothing more than the
three paragraph description I gave in my previous mail. Why should
implementors bother then?

This is what I (as a vendor) consider:

Should I implement RFC3195?

Pros
* It gives me an extendable transport protocol with integrity protection,
  authenticity, and negotiation
* It's standard and documented
* My current implementation has interoperability problems which would be
  solved

Cons
* I already have a protocol over TCP implemented which can be SSL wrapped
  (thus provides integrity protection and authenticity)
* I have no need for negotiation
* My current implementation interops with a couple of others, and it is
  quite easy to solve the interoperability problems with trivial
  modifications
* There are fewer RFC3195 compliant syslog peers than syslog/tcp capable
  syslog peers

So the decision to go for RFC3195 right now is not trivial. In the short
term ensuring interoperability between syslog/tcp implementations would buy
you more than implementing RFC3195. How about adding this paragraph to my
previous description:

NOTE: syslog/tcp is used in legacy syslog implementations and as such new
implementations should favor the protocol described in RFC3195.

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

Reply via email to