On Thu, Sep 08, 2005 at 03:30:28PM -0400, Sam Hartman wrote: > Note that the mandatory to implement solution for netconf is based on > ssh and does not support call home. 1) I see lots of people using tools like MIB browsers, snmp command line tools called from fancy scripts, monitoring packages such as cacti and nagios (these were even used on the IETF network in Paris if I recall correctly) and all these packages share the feature that the "manager" does not have a fixed transport address but takes the initiative to talk to an "agent" when there is need to do so. In other words, all these existing and deployed applications simply won't work with call home easily since they would require to implement some rather magic logic to dispatch SNMP traffic via a locally centralized connection manager. 2) It is important to talk about ssh and to not reduce the problem to just TCP. As far as I understand ssh, authentication is not symmetric because ssh has a clear buildin client/server role (servers authenticated via host keys while clients are authenticated via passwords or publickey mechanisms). So in the context of ssh, it does have quite some impact who establishes the transport connection. (Unless there is a "turn" feature in _ssh_. I am not aware of such a mechanism, but I am happy to learn.) 3) For those not following ISMS closely, it might be useful to know that originally was another proposal (EUSM) which tried to build on USM messages and tried to integrate USM with AAA infrastructures. This approach did have some rough support but did melt down because of some inappropriate use of EAP. (I can't explain the details since I never really understood the _technical_ concern.) Once EUSM was off the table, the WG converged to SSH for a very simple reason: SSH is already on many routers/bridges/hosts and you find it even on power strips these days. Netconf also requires SSH as the mandatory to implement transport. Hence, using SSH as a unifying mechanism to securely access the various management interfaces on a box has some promises wrt. ease of deployment. I agree with those who said that CH is an architectural change and I have yet to see a concrete proposal how CH via ssh can be achieved. /js -- Juergen Schoenwaelder International University Bremen <http://www.eecs.iu-bremen.de/> P.O. Box 750 561, 28725 Bremen, Germany _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf