I've had second thoughts. The data used for packet classification by
subscriber and realm is actually configuration data. "When in doubt,
throw it out": this data didn't belong in the MIB in the first place
unless someone sees it as particularly valuable for trouble-shooting.
Tom Taylor
On 04/12/2014 7:24 PM, Tom Taylor wrote:
I've been thinking through the implications of the NAT application
scenarios presented in my E-mail of 21 November ("NAT MIB: compliance
levels"). One of them is that the large-scale NAT application, even
though it does not understand individual subscribers, must have a means
to classify received packets into address realms. The conclusion is that
the packet classification criteria represented by
natSubscriberIdentifierType, natSubscriberVlanIdentifier, etc. in the
-11 version of the NAT MIB draft should be pulled out of
natSubscribersTable and put into a separate realm table. My instinct is
to make this NAT instance specific, in that:
-- different NAT instances on the same device may support different realms
-- different NAT instances may use different criteria to distinguish the
same realms, as a kind of load-sharing control
Comments?
Tom Taylor
_______________________________________________
OPSAWG mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/opsawg
