Re: [tsvwg] Robert Wilton's No Objection on draft-ietf-tsvwg-datagram-plpmtud-19: (with COMMENT)

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

Hi, Jonathan,

On Fri, Apr 10, 2020 at 9:02 AM Jonathan Morton <[e-mail]>
wrote:

> > On 10 Apr, 2020, at 4:38 pm, Spencer Dawkins at IETF <
> [e-mail]> wrote:
> >
> > ISTM that asking people to poke at this and report back in MAPRG is an
> excellent idea. Depending on whether there are still "relatively few MTU
> values in use" or not, it might then be useful to consider either updating
> the table of commonly-used values, or saying "binary search will probably
> work fine now".
>
> In connection with this, it's perhaps worth noting that some applications
> won't need to establish the precise MTU, only the minimum number of
> fragments they need to use for their jumbo datagram.
>
> For example, a 10,000 byte IPv4 packet can fit through a 1500-byte MTU
> path in seven 1446-byte fragments, but not in six 1684-byte fragments.
> (The calculation is different for IPv6 due to the larger headers.)
> Refining this distinction further only becomes relevant when datagrams of
> different sizes have to be sent over the same path, or when a stream
> transport is involved so that testing specific divisors is not helpful.
>

Thank you for pointing that out! I've also been ignoring jumbo frames,
which weren't a thing in 1990 because we were still trying to get Ethernet
to work well over coax, and I suppose there'd be some interaction there,
too.


> So a variety of search algorithms may be useful in different circumstances.
>

Yup. I hope the next stop is in MAPRG, because this does seem useful.

Best,

Spencer

>  - Jonathan Morton