Re: [tcpm] Early assignment of IANA TCP option number for AccECN

[Michael Tuexen <[e-mail]>]

> On 25. Jul 2022, at 17:31, Bob Briscoe <[e-mail]> wrote:
> 
> tcpm, Michael,
> 
> On 25/07/2022 15:08, Scheffenegger, Richard wrote:
>> 
>> 
>> Am 25.07.2022 um 16:02 schrieb Michael Tuexen:
>>>> On 25. Jul 2022, at 15:52, Scheffenegger, Richard <[e-mail]> wrote:
>>>> 
>>>> 
>>>> Hi,
>>>> 
>>>> As one of the authors of the AccECN draft, which approaches the final
>>>> stages, I wanted to see if there is support to ask IANA for an early
>>>> assignment of the two TCP Option numbers for AccECN.
>>>> 
>>>> The background here is, that the current use of the experimental option
>>>> in Linux (0xACC0 and 0xACC1) is not really documented - the draft only
>>>> mentiones the experimental ID 0xACCE which was in use prior of the
>>>> decision to go with two distinct option numbers.
>>> If I understand the experimental options correctly, you could just use
>>> kind=253, ExID=0xACCE
>>> kind=254, ExID=0xACCE
>>> to distinguish your two option using the assignment already made (for
>>> example the kind=253 for Order 0, kind=254 for Order 1).
>> 
>> Yes, could have :)
>> 
>>> I don't see the ones used by Linux under
>>> https://www.iana.org/assignments/tcp-parameters/tcp-parameters.xhtml#tcp-exids 
>> 
>> No. I went to
>> 
>> https://github.com/L4STeam/linux/blob/testing/include/net/tcp.h
>> 
>> to find the currently in use experiment IDs as observed in the interop
>> testing.
> 
> Richard told me that earlier today he asked IANA to document the two exIDs at https://www.iana.org/assignments/tcp-parameters/tcp-parameters.xhtml#tcp-exids
> Which is why I also documented them in the draft update I just sent.
> However, IANA hasn't updated that registry yet, AFAICT.
It might take them a while.

But if someone uses exIDs for early deployments of an ID, the ID should
specify how the exIDs should be used... Any chance to do that? Or does
it not make any sense anymore?

Best regards
Michael
> 
> 
> Bob
> 
>> 
>>>> 
>>>> However, the implementations between Linux and FreeBSD deviated - Linux
>>>> using two different experimental IDs (see above), while the FreeBSD
>>>> variant differentiated by an additional byte after the experiment ID
>>>> (0xACCE).
>>> I think the draft should describe how to use the experimental options.
>>> 
>>> Best regards
>>> Michael (as an individual)
> 
> -- 
> ________________________________________________________________
> Bob Briscoe                               http://bobbriscoe.net/
>