Re: [IPFIX] IPFIX export: SNMP versus physical Interface

[Juergen Quittek <[e-mail]>]

Hi Benoit,

I think we cannot change the specification of an IE once it has been
registered at IANA.
If we do so, existing implementations based on the spec at IANA may break.

    Juergen


On 26.07.11 16:29, "Benoit Claise" <[e-mail]> wrote:

>Juergen,
>
>> On Tue, Jul 26, 2011 at 10:54:52AM -0400, Andrew Feren wrote:
>>
>>>  From a collection and reporting perspective the important difference
>>>for
>>> me in these two sets of definitions isn't physical vs maybe not
>>> physical, but that "The value matches the value of managed object
>>> 'ifIndex' as defined in RFC 2863 ..." and everything that implies for
>>> elements 10/14.
>>>
>>> So I agree with Paul's approach.  There are situations where either
>>> 10/14 or 252/253 could be used, but if RFC 2863 applies to your
>>> interface use 10/14.
>>>
>>> Maybe rather than replacing "IP interface" with "logical or virtual
>>> interface" it could just be "interface" or "network interface".
>>>
>>> On a related not I'm not sure the references for 252/253 need to refer
>>> readers to "[RFC2863] for the definition of the ifIndex object."
>> RFC2863 always applies to an ifIndex object. I fail to see the
>> problem. I find the definitions of these elements rather clear as they
>> are written.
>I agree.
>Both sets (10/14 and 252/253( are ifIndex.
>The second set applies to the physical ifIndex, while the first one
>might be non-physical.
>A Flow Record can contain both, and the Collector can draw the
>conclusions if the sets are different.
>
>So I would not change the definitions.
>A clarification might be welcome (even though I'm not convinced).
>However, where?
>- not in IANA, which specifies individual IE, and not the relation
>between them
>- maybe in the RFC5102 Proposed Draft RFC
>
>Regards, Benoit.
>>
>> /js
>>
>
>_______________________________________________
>IPFIX mailing list
>[e-mail]
>https://www.ietf.org/mailman/listinfo/ipfix