Re: [ippm] WGLC for draft-ietf-ippm-delay-var-as.

[Al Morton <[e-mail]>]

At 05:05 AM 10/9/2008, Benoit Claise wrote:
>>Rudiger wrote:
>>I'd be interested in learning whether Henk's and Matt's implementations
>>apply this definition or another one. The question goes to all other
>>implementiors too. The more natural selection function seems to be to
>>evaluate pairs of consecutively received packets ignoring their
>>sequence numbers. Draft-ietf-ippm-delay-var-as however works with an
>>IPDV definition assuming awareness of sequence numbers. I'd like to
>>learn whether there are any implementations following the IPDV
>>definition of draft-ietf-ippm-delay-var-as (see section 6.1).
>>
>The Cisco IP SLA feature follows the IPDV definition
>
>Regards, Benoit.

Hi Ruediger,

First, thanks for finding a few typos!

The specific case of packet loss in section 6.1 is also
mentioned in section 4.1 of RFC 3393:
http://tools.ietf.org/html/rfc3393#section-4.1
(don't confuse the selection function with the
conditioned statistics described here, they are
different processes).

In my experience, selection functions are primarily applied
at the Source, and the RFC 3393 definitions refer to index numbers
(or source sequence numbers).  A selection function like you describe
("pairs of consecutively received packets ignoring their
sequence numbers") is certainly possible, and it's important
to understand how different things can be when comparing
results from two different selection functions.
For example, some of your observations on re-routes may not hold for
the IPDV defined in the memo - lost packets and undefined pairs
are a key factor when trying to detect path changes with
"consecutive source packet" IPDV, as we explained.

The WIPM measurement system described in
http://tools.ietf.org/html/draft-ietf-ippm-implement-02
(expired I-D, but I'd still like to see this published as an
Informational RFC, even if it's not a formal part of the
standards track progress of the original RFCs)
implemented both IPDV and PDV as described in the memo.

thanks again, and regards,
Al

_______________________________________________
ippm mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/ippm