Re: I-D Action:draft-ietf-v6ops-ipv6-cpe-router-03.txt

[Ole Troan <ot@cisco.com>]

Konrad,

>> it is a "MAY" and a "hint".
>> a typical scenario would for example be that an SP gave a /56 regardless of
>> the hint.
> 
> I hope /48 is more typical... ;-)

sure, that's up to the RIRs and SP policy.

>> does the working group feel a need to specify the case where the CE router
>> doesn't get a large enough prefix delegated? or are we happy leaving that
>> up to implementations. alternatives are at least "various degrees of
>> bailing out", "NAT66", "proxy ND"...
> 
> I personally feel this should be specified. Or at the very least some 
> acceptable options should be listed.
> 
> Bailing out: does this mean fe80::/10 only? Or fc00::/7 ?
> Note: fe80::/10 is a major headache  for users - I don't think I can explain 
> to most of my friends how to enter http://[fe80::1122:33ff:fe44:5566%1] into 
> their browsers and why it is different for every CPE they use.
> 
> NAT66: first off, the draft expired. Second it does not solve the problem: it 
> is not a cone NAT - it translates a /48 into another /48 (it might be adapted 
> to translate /n into /n, but not into a cone NAT).
> 
> What do you mean with "proxy ND"?
> 
> For most users a web proxy might be a temporary solution to deal with a no 
> delegation situation.

I added the other options partly tongue in cheek. the form of bailing out I had in mind was to not enable IPv6 on interfaces which the CE doesn't have addresses for. and raise the alarm.

> Question aside from this: why request only a prefix that is barely big enough 
> for a /64 for each interface? There could be dozens of routers behind that 
> CPE. Why not encourage to request a /48 everytime? We aren't running out of 
> them any time soon.

it is not quite a request. it is just a hint. as such another good option is to remove the text about the hint altogether from this draft.

Best regards,
Ole