From: Iljitsch van Beijnum [mailto:iljitsch@xxxxxxxxx] Sent: Tuesday, September 06, 2005 2:22 PM To: Eliot Lear >>> The IETF has been doing extensive work on NAT traversal, have a look >>> and see if you can reuse some existing mechanism. >> All mechanisms used with the possible exception of an additional SNMP >> table will be re-used from existing IETF work (mostly SSH with help >> from >> the fact that it's based on TCP). >You do realize that you import all the weaknesses of TCP then, don't >you? This issue was discussed within the ISMS WG over a period of several weeks. During those discussions, I tried to explain why I believe that a UDP transport alternative for ISMS needs to be available. I offered the following technical reasons supporting this position: 1) Our experience is that SNMP over TCP has significantly worse behavior in bandwidth constrained environments (e.g., wireless links) and during times of network congestion than SNMP over UDP. 2)TCP's inappropriate behavior (i.e., treating latency as congestion and thus "backing off") in high latency environments such as across geo-stationary satellite links. I do not understand why the WG did not resonate with these observations, since these are very well known issues and the technical underlying reasons easy to understand. I can only conclude that while these issues are very important to the deployments with which I am familiar, they are sadly not relevant to the majority of the ISMS WG. I believe that ISMS would be a better solution if it could operate within all envisioned environments. --Eric _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf