[IPv6]Re: Gorry Fairhurst's Discuss on draft-ietf-6man-eh-limits-19: (with DISCUSS and COMMENT)

[David Farmer <[e-mail]>]

On Mon, Apr 28, 2025 at 10:29 AM Tom Herbert <tom=
[e-mail]> wrote:

> On Mon, Apr 28, 2025 at 8:06 AM Michael Sweet <[e-mail]> wrote:
> >
> > All,
> >
> > My $0.02 CAD: If currently deployed hardware supports 64 byte EHs at 80%
> (for whatever reason) and currently sold hardware supports 256 byte EHs,
> why not set an aspirational requirement in eh-limits of "MUST support a
> minimum of 128 byte EHs" and provide an informative reference to the paper
> showing the limits of Internet-wide EH support data at time of publication?
>
> Michael,
>
> Aspirations are great, but that's all they are. That doesn't help an
> application developer trying to use Extension Headers today, not some
> time in the far future when the whole Internet has been retrofitted
> with the latest hardware. Even going to 128 bytes has a significantly
> higher loss rate than 64 bytes (>60%). IMO it is the purview of IETF
> to give guidance to developers. Vague guidance, like " one must pay
> attention to its size" (from the paper) or "A Source MAY .. limit
> their size to increase the likelihood of successful transfer across a
> network path" (RFC9673) isn't useful to an application developer
> building an app that is expected to run across the whole Internet. We
> either need a quantified recommendation or just continue to not use
> Extension Headers.
>
> Tom
>

I believe you are all correct; I agree with Tom, the data shows that
today's practical or usable limit is 64 bytes. Nevertheless, I also agree
with Tim and others that we don't want hardware manufacturers to converge
on 64 or 128 bytes; I believe the goal should be at least 256 bytes. But we
have to be clear that it is an aspirational goal, not today's
practical limit, which is clearly 64 bytes.  I also agree with Tom; we are
where we are because we haven't set specific limits or goals. As a network
operator, I want EHs to work, but I have nothing to measure vendors
against. I need something with a specific number to push my vendors for.

This doesn't necessarily need to be an IETF number, but we need a number.
My example, around 25 years ago, the Research and Education (R&E)
networking community set 9000 bytes as the Ethernet MTU goal for our
community. This has been mostly successful; the vast majority of the R&E
Networking community supports a 9000-byte Ethernet MTU.

Thanks.

-- 
===============================================
David Farmer               Email:[e-mail]
Networking & Telecommunication Services
Office of Information Technology
University of Minnesota
2218 University Ave SE        Phone: [phone]
Minneapolis, MN 55414-3029   Cell: 612-812-9952
===============================================