Re: I-D ACTION:draft-hardaker-snmp-session-sm-00.txt

>>>>> On Tue, 14 Oct 2003 08:01:51 -0500, "Golovinsky, Eugene" <[email protected]> said:

Eugene> My understanding is that when sessions die from natural causes
Eugene> it'll clean up key(s) associated with it. That is done without
Eugene> MIB involved. On the other hand I would like to be able to
Eugene> manage and administer the entity while session is active (per
Eugene> session statistics, retries, failed sessions, whatever else)
Eugene> and to be able to cancel the key if there is a need. If not
Eugene> through the MIB, then we'll have to come up with other way
Eugene> (protocol ??) to do that. I'd rather stick to SNMP here.

My expectation is that the following three methods will be available
for tearing down connections:

1) An in-protocol CONNECTION-DELETED to inform the other side that it
   should release and clean resources (MUST actually, especially
   keys).  Functionally, it's a FIN.

2) A MIB method for control by SNMP, either within the same session or
   by someone with administrative rights.

3) If either side simply deletes its side of the session, then the
   other side will be unable to communicate over it (this must always
   be supported since things like reboots will normally clean the
   session information within a device).

-- 
Wes Hardaker
Sparta