Re: Manual PMTUD [was ...rfc2460bis-08]

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

On 18/03/2017 22:15, Robert Raszuk wrote:
> Hi Brian,
> 
> On Sat, Mar 18, 2017 at 2:10 AM, Brian E Carpenter <
> [e-mail]> wrote:
> 
>>> I can use
>>> today SRv6 features in my data centers where I consistently know that all
>>> interfaces are set for jumbo frames and where hopefully I do not need to
>>> worry about MTU since hosts have much lower max MTU configured.
>>
>> Exactly. That's not the Internet. Internet Standards are for the Internet,
>> hence I believe that the AD's decision is the correct one.
>>
> 
> 
> ​Well I don't know .. Protocols are used both in the open Internet and in
> the contained ​environments. Hence IANA allocations and draft progress
> should be encouraged regardless if something is to be deployed in the
> Internet or not.

That's true, and that's why I'm very open to the idea of local-use
allocations in general. But for the IP protocol above all, we need
Internet-wide interoperability in the Internet Standard.

    Brian

> 
> Now as customer I do encourage multi-vendor networks even for closed
> RFC1918 private WAN, MAN or DCs.
> 
> If we just drop and position IETF to work on "Internet" only use cases
> where would interoperability of routing protocols and beyond fall under ?
> Which SDO ?
> 
> 
> (How would you feel about an extension header that MAY be deleted
>> by an intermediate node, and MUST be deleted if the next hop's link
>> MTU is too small?
> 
> 
> 
> ​Well SRH can be applied both as a mission critical or optimization. If it
> is an optimisation ​indicated in the packet by a flag I would be fine to
> remove it if it does not fit to the MTU of the link.
> 
> If it is however "MUST HAVE" for a given flow I would rather see a packet
> drop then to continue forwarding without it.
> 
> Both such events should indeed be signaled to the "owner" of the SRH or any
> other extension header either inbound via for example ICMPv6 or maybe by
> some other targeted message or indicated out of band (via logging to
> syslog) - applicable to single administration domain.
> 
> ​Kind regards,
> Robert.​
>