Re: What to do about multipath in QUIC

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

Hi, Martin,

On Thu, Nov 5, 2020 at 2:59 PM Martin Duke <[e-mail]> wrote:

> Hi Spencer,
>
> I shy away from deciding "how multipath should work." I think we should
> figure how much protocol we need for the experimenters to figure out how
> multipath should work. But maybe that's what you meant.
>

If I didn't mean that, I should have (especially after taking a closer look
at https://datatracker.ietf.org/doc/draft-huitema-quic-mpath-option/ today.
I think "how much protocol we need to figure out how multipath could work"
is a great revision to my third second bullet point.

Thanks for asking about that.

Best,

Spencer


> On Wed, Nov 4, 2020 at 3:39 PM Spencer Dawkins at IETF <
> [e-mail]> wrote:
>
>> Lars said that if we wanted to move forward on multipath in QUIC, we
>> should be talking about multipath in QUIC on the mailing list, so this is
>> my suggestion as a starting point, based on my understanding of where we
>> ended up after the October virtual interim meeting and resulting threads on
>> the mailing list.
>>
>> It's pretty clear that a lot of people have deploying and testing
>> connection migration as their priority in the near future, and we should
>> not distract them from that worthy task.
>>
>> It's clear that at least some people think that connection migration onto
>> a new connection that has already been validated is a lightweight
>> operation. Deploying and testing connection migration will be a good basis
>> for verifying that theory.
>>
>> It's clear that there are different ways of thinking about multipath in
>> QUIC - https://datatracker.ietf.org/doc/draft-deconinck-quic-multipath/
>> is the proposal I have the most experience with, but Yanmei was presenting
>> https://datatracker.ietf.org/doc/draft-an-multipath-quic/ at the virtual
>> interim, and Christian submitted
>> https://datatracker.ietf.org/doc/draft-huitema-quic-mpath-option/ a day
>> or two before the virtual interim. That leads me to make two suggestions:
>>
>>    - That we get experience with the proposals that we already have, and
>>    any proposals that pop up in the meantime.
>>    - That we discuss that experience and work on coming to a consensus
>>    about how multipath should work before moving forward.
>>    - That we publish (one or more) multipath proposals as Experimental,
>>    if and when that's the right thing to do.
>>
>> Thoughts?
>>
>> Best,
>>
>> Spencer
>>
>>