Re: [tcmtf] BoF feedback

[Spencer Dawkins <[e-mail]>]

On 8/6/2013 3:00 AM, Jose Saldana wrote:
> If you look at the abstract of the "Recommendations draft"
> (http://datatracker.ietf.org/doc/draft-suznjevic-tsvwg-mtd-tcmtf/) , this is
> exactly what we intend to do in that document:
>
>     This document contains recommendations to be taken into account when
>     using methods which optimize bandwidth utilization through
>     compression, multiplexing, and tunneling traffic flows (TCMTF) over a
>     network path.  Different multiplexing policies and implementation
>     issues which are service and link specific are discussed.
>     Additionally, this document describes *policies which can be used for
>     detecting, classifying, and choosing the network flows suitable for
>     optimization by using TCMTF*.  Finally, recommendations of maximum
>     tolerable delays to be added by optimization techniques are reported.
>     Recommendations are presented only for network services for which
>     such bandwidth optimization techniques are applicable (i.e., services
>     with low payload to header size ratio, which will also be denoted as
>     "small-packet flows").
>
> Spencer, are you talking about the "policies which can be used for
> detecting, classifying, and choosing the network flows suitable for
> optimization by using TCMTF"?

I think so. And it's possible that a word choice is confusing me. 
"detecting" sounded pretty on-the-fly/dynamic to me (as in, you might 
detect that any random packet going past would be a packet that TCM-TF 
would process), but my understanding of recent e-mails is that 
"detecting" is much less dynamic ("I want to use TCMTF on WoW, and 
here's how you would know a packet is a WoW packet, and anything I 
didn't say how to recognize as a packet that TCM-TF should process is 
forwarded with no changes").

If my understanding based on these recent e-mails is more in line with 
what you're thinking, that answers my question.

Thanks!

Spencer

> Perhaps this was not clear enough in the BoF, but of course this is very
> important: you can only use TCM-TF for the TCM-suitable flows, and this spec
> must clearly define which flows are TCM-suitable and in what conditions.
>
> Thanks,
>
> Jose