<inline>
Tom Petch
----- Original Message -----
From: "Daniel Senie" <dts(_at_)senie(_dot_)com>
To: "Juergen Quittek" <quittek(_at_)netlab(_dot_)nec(_dot_)de>
Cc: "IETF Discussion" <ietf(_at_)ietf(_dot_)org>
Sent: Thursday, September 08, 2005 4:10 PM
Subject: Re: ISMS working group and charter problems
At 09:14 AM 9/8/2005, Juergen Quittek wrote:
--On 9/7/2005 6:49 PM -0400 Sam Hartman wrote:
"Fleischman," == Fleischman, Eric
<eric(_dot_)fleischman(_at_)boeing(_dot_)com> writes:
Fleischman,> I believe that network management is too important a
Fleischman,> functionality to be designed such that it can only be
Fleischman,> usable within highly confined environmental
Fleischman,> constraints.
"must work everywhere," is a highly constraining environment.
We should consider that ISMS is about integrating SNMP into user
and key management systems. Such system usually operate over TCP.
You mean like RADIUS? That's a UDP protocol.
In a highly damaged network ISMS might not be able to help you even
if you had stuck to SNMP transport over UDP.
So you're arguing we should not bother with the ISMS effort?
You certainly may revisit the UDP vs TCP decision on the IETF list;
doing so is an appropriate last recourse under our process.
However I do not believe it likely that you will get IETF consensus on
a specific UDP direction. I also do not believe it would be
productive to take this issue back to the working group. So, I ask
what you believe I should do if you fail to get consensus? If your
options are no ISMS or ISMS over ssh, which would you pick?
The consensus in the WG on this issue was not really rough,
but rather broad and clear.
Based on your email, the consensus of the group is that TCP is good
enough, since it'll only be interesting to manage networks that are
operating cleanly. I can't imagine that's what the WG really
concluded, but that's how your email reads.
As a WG member, the way I saw it was that TCP became a necessary evil; the
choice of a secure transport - mutual authentication, message integrity,
encryption - came down to TLS+SASL, DTLS+SASL, SSH. Among these, SSH dominates
in the market place of operators so, for me at least, it was a one horse race.
Then having to have TCP, with all its complications is, well, a necessary evil
(but one that the market place lives with at least for secure remote login).
There is no, and no sign of, a SSH over UDP nor of SSH over any more suitable
transport. So SNMP over TCP in order to get integrated security is a leap in
the dark. It could work, it could fail, it may never even make it to an RFC,
but it offers the best if not the only hope of progress.
The issue of TCP in an unreliable network did get discussed several times
without any consensus. For me, if isms allows safe configuration changes in a
working, reliable network without requiring an SNMP-specific security subsytem,
then that will be such a giant leap forward for SNMP that it will be worth the
effort. Other WG members want more, secure notifications, Informs, call home
and such like; for me, those can wait.
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf