Re: I-D Action: draft-ietf-6man-ipv6only-flag-03.txt

[Ole Troan <[e-mail]>]

Hi Nick,

[deleting all text apart from the technical point]

> Here's an example.  Drafts -00 to -02 contained the following:
> 
>>   o  In particular, this may overload switches in multi-segment
>>      wireless networks because it will create IPv4 state for every dual
>>      stack host.
> 
> I posted an analysis on Sep 8 (https://mailarchive.ietf.org/arch/msg/ipv6/Yoy1Bo8A6WiRryqogTfr_R7FH9I) which went into some nuances about how switches work in the context of forwarding ipv4 frames, but the summary was:
> 
> "...layer 2 switches do not carry ipv4 state in any meaningful sense. It is technically misleading, bordering on flat-out incorrect, to suggest otherwise."
> 
> The draft was then updated from:
> 
> > this may overload switches [...] because it will create IPv4 state
> 
> to:
> 
> > this may overload switches [...] if the switches create IPv4 state
> 
> I'm at a loss as to how to respond to this other than to say that switches don't create ipv4 state and that it is disturbing to have to explicitly state something so fundamental in an IETF working group, particularly in the context of statements like: "the authors think that all your points are in fact dealt with in the text”.


Firstly, this seems to be background information, provided in the draft to justify the solution.
I say this to distinguish the point from the real technical issues with the protocol mechanism itself.

To your point. You are right of course. Maybe the authors have conflated bridges and routers? And that they are really talking about a router (aka L3 switch) that has limited TCAM space for forwarding state? Regardless I think this is a weak point, given that IPv6 in those terms use far more state. Easily an order of magnitude.

If no-one else in the working group feels strongly for this point, I would suggest we simply remove that bullet.
Or make a very vague statement about "disabling IPv4 completely on a link, ensures that no IPv4 state is required in the network or in hosts."

Best regards,
Ole