Re: Spin bit decision

[Mikkel Fahnøe Jørgensen <[e-mail]>]

Are there any concerns about impact and how to handle or ignore malicious
manipulation of the spin either by endpoint or a middlebox?



On 2 October 2018 at 14.56.16, Lars Eggert ([e-mail]) wrote:

Hi,

again, as an individual:

On 2018-10-2, at 15:41, <[e-mail]> <
[e-mail]> wrote:
> Well, if the spec says that en endpoint MUST copy of flip a bit, an
implementor would be fairly ill-advised not do comply.

sure, it may be "ill-advised", but in practice, it won't affect interop at
all.

> And nothing prevents from adding a "does it spin ?" criterion to the
interop test suite.

We can certainly consider this - it may give us some data about the current
capabilities and configurations of the participating stacks.

> Clearly you now have this degree of freedom: weaken that MUST into a MAY
or a SHOULD, separately for each role. But for what purpose ?

The point I'm trying to make is that it in practice doesn't matter if the
spec says "MUST", "SHOULD" or "MAY", since it doesn't affect interop, and
doesn't affect performance. Essentially, each stack can decide in isolation
whether to spin, echo, do nothing, grease the bits, etc., no matter what
language we put in the spec.

Lars