Re: [LOOPS] [nwcrg] BOF co-chairs thinking on LOOPS next steps

[Spencer Dawkins at IETF <[e-mail]>]

Hi, Colin,

On Wed, Jul 31, 2019, 07:36 Colin Perkins <[e-mail]> wrote:

> On 30 Jul 2019, at 18:43, Spencer Dawkins at IETF <
> [e-mail]> wrote:
>
> I'm going to be replying to two or three e-mail in this thread, so I'll
> try to be coherent over my replies ....
>
> On Tue, Jul 30, 2019 at 9:42 AM Marie-Jose Montpetit <
> [e-mail]> wrote:
>
>> Since most of your questions are related to FEC I copy the nwcrg. See
>> <mjm> below
>>
>> mjm
>>
>> Marie-José Montpetit, Ph.D.
>> Research Affiliate, MIT Media Laboratory
>> [e-mail]
>> [e-mail]
>>
>> On July 30, 2019 at 10:23:04 AM, Spencer Dawkins at IETF (
>> [e-mail]) wrote:
>>
>> Hi, Mirja,
>>
>> On Tue, Jul 30, 2019 at 5:34 AM Mirja Kuehlewind <[e-mail]>
>> wrote:
>>
>>> Hi Carsten, hi Spencer,
>>>
>>> > On 25. Jul 2019, at 20:59, Spencer Dawkins at IETF <
>>> [e-mail]> wrote:
>>> >
>>> > My impression, and Magnus might say I'm insane, was that Magnus is
>>> right on the edge of whether this work can be chartered without a second,
>>> working group-forming BOF. If the LOOPS community can't live without
>>> host-to-node, the community should add it. If the community can make use of
>>> LOOPS without host-to-node and add it later, the community might consider
>>> whether that reduces the scope of the initial proposed charter enough to
>>> help Magnus approve it :-)
>>>
>>> Yes, it would be possible to charter without another BoF a small and
>>> well-scoped group, however, it also would be possible to work on small and
>>> well-defined pieces of work without a new working group. This was one of
>>> the reasons why this was not a non-wg-forming BoF because it is not clear
>>> yet if a new group is needed or the resulting pieces of standardization
>>> work are small enough to fit into an existing group.
>>>
>>> Therefore, I really recommend focus on defining the standardising work
>>> items that you think are needed rather than trying to do any word-smithing
>>> of a potential charter (that eventually is not needed).
>>>
>>
>> Thanks for giving me the chance to clarify - I really was talking about
>> chartering work, not just chartering a working group, but that wasn't clear
>> from my note. As you said, knowing what the work is, in more detail, will
>> help you and Magnus know what to do next.
>>
>> So, my suggestion for the interested community is to nail down
>>
>>    1. In order to "do LOOPS", what already exists, that can be used
>>    without changes?
>>    2. what already exists, but needs to be extended for LOOPS?
>>    3. what needs to be created, because nothing exists that meets the
>>    needs?
>>
>>  I'm not a LOOPS proponent (the ADs asked me to chair the BOF about a
>> month before IETF 105), but speaking as someone who hasn't been involved in
>> depth, I wonder about
>>
>>    1. How a sender tunes FEC dynamically - is that automatic, based on
>>    FEC mechanisms people are thinking about, or is there work to do there?
>>
>> <mjm> There are FEC mechanisms to adapt coding dynamically based on
>> acknowledgments for example. In the QUICK implementation we will have flag
>> a RECOVERED packet that could be used to adapt the coding rate. Other
>> implementations or code-specific protocols are also available.
>>
>>
>>    1.
>>    How a sender knows whether to do FEC, retransmission, or both FEC and
>>    retransmission, dynamically?
>>
>> <mjm> Well the use of FEC (or not) is a design decision.  Of course you
>> need the code (and the libraries). And then the decision will be based on
>> known link condition, goodput, expected performance, type of traffic
>> (video, time sensitive etc.). And finding a tunnel and PEPs if you need to.
>> And it can depend on path statistics and not being end to end. I am not
>> sure you could standardize that decision.
>>
>
> So, speaking with no dots because I don't have any, I agree with what
> Marie-Jose is saying here, but my understanding is that LOOPS endpoints
> might turn LOOPS off completely if local optimizations aren't needed. So
> I'm thinking about how a sender makes that decision, and how a sender might
> do trade-offs.
>
> I think one point in discussions with Marie-Jose might be that she's
> thinking in terms of Technical Specifications (
> https://tools.ietf.org/html/rfc2026#section-3.1), which are necessary,
> but discussions are making me think that proponents might usefully also
> think in terms of Applicability Statements (
> https://tools.ietf.org/html/rfc2026#section-3.2), which (quoting)
>
>    An Applicability Statement specifies how, and under what
>    circumstances, one or more TSs may be applied to support a particular
>    Internet capability.  An AS may specify uses for TSs that are not
>    Internet Standards, as discussed in Section 7.
>
>    An AS identifies the relevant TSs and the specific way in which they
>    are to be combined, and may also specify particular values or ranges
>    of TS parameters or subfunctions of a TS protocol that must be
>    implemented..  An AS also specifies the circumstances in which the use
>    of a particular TS is required, recommended, or elective (see section
>    3.3).
>
>    An AS may describe particular methods of using a TS in a restricted
>    "domain of applicability", such as Internet routers, terminal
>    servers, Internet systems that interface to Ethernets, or datagram-
>    based database servers.
>
> So, if Applicability Statements (also standards-track documents) would be
> helpful for LOOPS deployment and operation, the proponents might think
> about whether they want to do work in that space.
>
> If decisions about what senders do turn out to be entirely up to the
> sender, that's great, but if the proponents want to bound the possible
> behaviors of implementations to improve deployment and operation, that
> might also be appropriate in BCPs (
> https://tools.ietf.org/html/rfc2026#section-5).
>
>
>>
>>    1.
>>    How a sender knows that it shouldn't be doing anything, because
>>    anything it does won't help ("first, do no harm")?
>>
>> <mjm> Of course in a congested network doing nothing is probably best.
>> Again there is a lot of work on the interaction of coding and congestion
>> control.
>>
>> <mjm> Not everyone knows the things but there has been a lot of work in
>> FECFRAME (in IETF) and NWCRG. Someone did mention that a lot of the work
>> was done in a research group not a working group. But real implementations
>> exist in famous “networks".
>>
> Right - so one possibility is publishing work from NWCRG in the IETF
> stream, and especially standards-track, with appropriate review by the IETF
> community. That could take a lot of forms (RGs and WGs can interact in a
> lot of different ways).
>
>
> An IETF WG could have a work item to develop a standards track FEC scheme
> based on the research previously done in NWCRG, certainly.
>
> <mjm> Since most of your questions are about erasure coding (coding for
>> packet loss) is that a direction LOOPS wants to take? Somekind of nwcWG?
>>
>
> See Mirja's note about chartering work, and only chartering working groups
> when that's needed. But as you note, the FECFRAME extensions happened in
> TSVWG, so that would be a reasonable conversation to have with TSVWG chairs
> (and I note that Wes Eddy, who popped up further down in this thread, is a
> TSVWG co-chair). If it fits, TSVWG could be a plan. If it doesn't fit in an
> existing working group, NWCWG could be a plan.
>
>
> I perhaps misunderstood LOOPS, but I’d assumed that – to the extent is
> uses FEC and not retransmission – it’d pull in one of the schemes developed
> in FECFRAME. Has there been any requirements analysis to show if anything
> further is needed?
>

That's the analysis that I'm suggesting the proponents work on now. I
didn't talk to all the proponents in Montreal, but at least a couple were
pretty sure they wanted to use FEC, but were still figuring out details on
that usage.

Marie-Jose and some other nice people from NWCRG are providing pointers to
help with that analysis. For which, I'm grateful. I was the responsible AD
for the recent sliding window FECFRAME extensions that migrated through
TSVWG, and I understand FECFRAME at a 10,000-meter level, but the
proponents (and interested parties) need to work through the next level of
details.

This is separate to the question of whether there's work in NWCRG that’s
> ready to move into IETF for standardisation, of course.
>

The nice thing about the existing relationship between NWCRG and TSVWG is
that if there's more stuff from NWCRG that's ready to standardize in the
IETF, we know how to do that, and Wes Eddy (one of the TSVWG co-chairs) is
already helping with congestion discussion on this mailing list.

I think the non-working group-forming BOF did a good job at flushing out
the questions and the resources needed to answer them. Thanks, Magnus and
Mirja, for making that easy.

Spencer

Colin
>
>
>
>
> --
> Colin Perkins
> https://csperkins.org/
>
>
>
>
>