Re: UDP source ports for HTTP/3 and QUIC

[Töma Gavrichenkov <[e-mail]>]

Peace,

On Sat, Jul 17, 2021, 3:28 AM Mark Nottingham <[e-mail]> wrote:

> I appreciate the enthusiasm for the tangential here :) but to return the
> original topic -- I realised that this is obliquely discussed in the
> applicability draft:
>
> https://quicwg.org/ops-drafts/draft-ietf-quic-applicability.html#name-port-selection-and-applicat
>
> So at a minimum, it seems like we should expand upon that text to make
> this issue more clear. I'll try to start a PR.
>
> If I read the thread correctly, people seem to think (somewhat
> pessimistically, but realistically) that protocols susceptible to
> reflection will continue to be created, so hardcoding a list into the RFC
> isn't workable.
>
> Instead, I'm currently thinking the best approach will be to:
>
> 1. Expand the applicability document as per above, using examples from the
> list I gave, since they're already pervasive
> 2. Start a discussion in the TSVWG about the possibility of adding a new
> column to the port registry to capture this information
>

What if we add a new key to the DNS SVCB record so that an HTTP/3-enabled
service could indicate the source port ranges it does [not] want or is
[not] able to see?

The HTTP/3 servers would probably be using SVCB anyway for a number of
applications.

--
Töma