Discussions on EOS at the SF IETF

Some of the items discussed in the informal EOS meeting at the SF IETF :

Strengthen the wording in the draft-ietf-eos-oops-00.txt to explain how the
cursor in the new pdus can be effectively used to walk a mib similar to
get-next.

There were questions on whether existing management applications would be
affected by these new operations. Since the data returned by the new
operations are not ordered the same way as done by the get-next operation.
It was explained that the new proposals do not alter the get-next
processing in any way and hence the applications that use get-nexts will
still work the same way that it always used to. Only new or changed
applications that make use of the new operations will be affected.

There was also discussions about how long the cursor should remain valid.
The draft specified that the cursor should be valid at any time in the
future after it is created. It was agreed that the cost of implementing the
cursor to be valid at any time in the future was not worth the benefit
obtained from it. It was suggested that the cursor be kept active for a
fixed amount of time (5 minutes). That is, on every request to the command
responder, the cursor could be updated to the current time and if the
command responder receives a cursor that is more that 5 minutes old it
could consider the cursor to be invalid and process it accordingly.

There was also discussion about duplicate rows being present in the values
returned by the command responder in rare circumstances when using the
cursor.  I will send out a separate note with that problem.

Please let me know if something was missed.

/ Dinakaran