On 26 August 2010 00:47, Suresh Krishnan <[email protected]>wrote:

> Hi Woj,
>
>
>  I think this is the basis of our disagreement. There are no assumptions
> that this document is making. There are usage scenarios not involving the
> mechanism proposed in this document that are relevant to this issue you are
> bringing up.
>

Indeed, this seems to be the base of the current misunderstanding. Based on
the available information, the intent of the rs-mark mechanism is to use the
additional line-id information at/by the edge-router to selectively assign a
*different* IP prefix to each subscriber (where subscriber = one or more
IPv6 end-devices behind a given Line-id). This has been corroborated by
statements made on the thread.
Based on all this, one can currently conclude that that such end-devices
will *only* obtain valid IP addressing info *after* they send an RS, and
then receive a response by means of a unicast RA, not at any time before
before. The latter aspect has been confirmed by the presence of periodic
PIO-less multicast RAs in the scheme.
Hence, I come to ask: Suppose an end-device time-outs in sending RSes before
getting an RA in response. How does a context relying on the rs-mark scheme
deal with this case? If the answer is "it uses PIO-less multicast RAs", then
it does not appear to be enough (based on RFC 4861 not mandating hosts to
respond to RAs). Alternatively, what assumptions are you going by (eg. a
special type of IPv6 end-device? a plan to modify rfc4861? other?)? IMO such
assumptions need to be documented


> Let's say the mechanism proposed in this draft did not exist in either the
> IETF or BBF. Does the problem of the lost RSs exist? I contend that it does.
> What do you say? Unless we have a common view on this, I do not think we can
> agree on a resolution.
>

I certainly agree with you that RS losses and/or timeout issues are not
unique to the rs-mark usage context, but unlike in the traditional case,
where periodically sent multicast RAs would contain a PIO allowing an
end-device to acquire an address, here I don't see how recovery from such
issues takes place. The result appears to be end devices having no IPv6
connectivity, which I'm sure you agree is not a good thing.
In other words, the RS lost/timeout problem appears to be particularly
severe in its consequences when used with what the RS-mark mechanism says it
is intended for.

Thanks,
Woj.


>
> Thanks
> Suresh
>
>
>
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to