Re: [v6ops] TCP and IPV6_USE_MINMTU

[Brian E Carpenter <[e-mail]>]

On 19/10/2015 11:58, Mark Andrews wrote:
> In message <[e-mail]>, Brian E Carpenter writes:
>> I think that
>> (a) This draft should not be Informational, it should become a BCP.
>> (b) I suspect it belongs in v6ops rather than 6man, since it doesn't
>> change anything in the standards.
> 
> Isn't this similar to RFC6691 which was only informational?

IMHO that's an error and 6691 should be standards track, but I wasn't
involved in that discussion so there may be a reason.

> That said I don't care which w.g. processes this so long as it gets
> processed.  Similarly w.r.t. to the category.  Lets let the chairs
> decide which w.g.

Sure. I would argue for BCP because it really is an operational requirement
to get this fixed in all implementations.

   Brian

> 
> For v6ops this is
> https://datatracker.ietf.org/doc/draft-andrews-tcp-and-ipv6-use-minmtu/
> 
> Mark
> 
>> (c) I think it doesn't say quite enough about how this is a very serious
>> and recurrent operational problem which causes real damage to IPv6 deployment
>> .
>>
>> Also - it's not OK that POSIX hasn't yet picked up the advanced socket API
>> and that it isn't available in all programming languages with socket calls.
>> We, for some value of "we", need to campaign to get this fixed.
>>
>>     Brian
>>
>>
>> On 17/10/2015 11:56, Mark Andrews wrote:
>>> In message <[e-mail]>, Brian E Carpenter writes:
>>>> On 17/10/2015 09:31, Mark Andrews wrote:
>>>>>
>>>>> In message <20151016202620.GI1442@verdi>, John Leslie writes:
>>>>>> Mark Andrews <[e-mail]> wrote:
>>>>>>>
>>>>>>> Please see
>>>>>>> https://www.ietf.org/id/draft-andrews-tcp-and-ipv6-use-minmtu-00.txt
>>>>>>>
>>>>>>> It addresses what should be obviously done but isn't by many
>>>>>>> implementations.
>>>>>>
>>>>>>    Alas, this draft gives no clue _how_ to check IPV6_USE_MINMTU ...
>>>>>> (or, for that matter, how to set it).
>>>>>>
>>>>>>    One could reasonably guess that s/he should look through RFC3542 to
>>>>>> find how to check this; but RFC3542 contains no such string. (It does
>>>>>> contain the string IPV6_USE_MIN_MTU; but I don't feel justified in
>>>>>> guessing that far astream...
>>>>
>>>> Incidentally, it recently came to my attention that standard Python
>>>> doesn't support RFC3542. I don't know how widespread that issue is, but
>>>> it may be behind some of these broken MSS deployments.
>>>>
>>>>     Brian
>>>
>>> The entire advanced socket API is poorly supported because POSIX
>>> didn't pick it up.
>>>
>>> That said I've updated the document based on the feedback received
>>> to date.
>>>
>>> https://tools.ietf.org/html/draft-andrews-tcp-and-ipv6-use-minmtu-03
>>>
>>>>> Yes, I typo'd that, repeatedly.
>>>>>  
>>>>>>    Perhaps adding a citation to the specific section of a particular
>>>>>> RFC would help?
>>>>>>
>>>>>> --
>>>>>> John Leslie <[e-mail]>