Re: RS sending in draft-ietf-v6ops-ipv6-cpe-router-04

[Wojciech Dec <wdec@cisco.com>]

On 28/04/2010 12:58, "Konrad Rosenbaum" <konrad@silmor.de> wrote:

> Hi,
> 
> On Wed, April 28, 2010 08:47, Wojciech Dec wrote:
>> Updating RFC4861 is not the intent and there is no modification of the
>> protocol being proposed here. Citing from the begging of this thread:
>> 
>> RFC4861 states:
>> " A host sends Router Solicitations to the all-routers multicast
>>    address.  The IP source address is set to either one of the
>>    interface's unicast addresses or the unspecified address."
>> 
>> As such having Rses with a unicast address is already covered by the
>> protocol. The proposed change tightens the CPE's RS sending rule, which
>> combined with serialized DAD (or even optimistic DAD), makes sense.
> 
> From a programmers perspective I do not think that it would be realistic
> to tighten this specific requirement and expect people to comply. There
> aren't many router vendors out there who write the entire software stack
> of their routers - this is especially true for (relatively cheap) CPE
> equipment.
> 
> The RS/RA and DAD protocols are usually implemented deep inside the
> operating system kernel. I can't even find an option for Linux to switch
> between RA with LL and null source address.
> 
> So in order to make sure I comply with such a requirement I would have to
> dig deep into the kernel source and probably change it, while I would
> already be under a very tight schedule fighting against my hardware
> platform vendors development kit.

That is an implementation detail. One can well argue that the CPE spec goes
beyond vanilla router v6 stacks in a number of other areas, eg the handling
of the RA flags. The BBF spec goes into even finer detail, and IMO it would
be a bit naive to have the CPE spec just represent some status quo of v6
implementations; we have a CPE spec so that vendors can build to it CPEs and
other devices (CMTSes, BNGs, etc), and operators have a model reference of
the expected behaviour.
Functionally, as I said previously, it's important for the RS that is used
to authenticate a given user to contain an IP address that can be attributed
to that same user at the BNG. With serialized DAD and RS sending on the CPE,
I really do see it as a logical thing to define.

-Woj.

> 
> Right now is one of the very few moments that I'm happy I don't work for a
> CPE router manufacturer. ;-)
> 
> On a related note: I would expect ISP side equipment to work with most
> common systems out there without forcing significant or unintuitive
> changes on them. In this particular case: find a way to inspect the MAC
> address or line-ID instead of relying on the LL source address of the RA
> message.
> 
> 
>     Konrad
>