On 2010-09-24 09:42, Yiu L. Lee wrote: > I am ready to send the question to 6man. Before doing this, I got a question > for GI-DS-lite. The spec assumes that there is a single pt-to-pt tunnel from > the GW to AFTR. If we use Flow Label, the BRAS will have a pt-to-mp tunnel > from BRAS to AFTR (each CPE will have a dedicated IP-in-IPv6 tunnel). Will > this violate the spec?
Isn't the real question whether this will add an unreasonable amount of state to the AFTR? Brian > > > On 9/14/10 1:41 PM, "Frank Brockners (fbrockne)" <[email protected]> wrote: > >> Hi Yiu, >> >>>> ...FB: Well, the concern is a more general one. I'm not disagreeing >>> that we >>>> might be able to have flow policies even with the new use of the >> flow >>> label. >>>> The thing that I think requires more discussion is whether people >> see >>> a need >>>> for using the IPv6 flow label in the core of their network, even if >>> GI-DS-lite >>>> with this new encapsulation would be used within the very same >>> deployment. If >>>> there is such a use, the new encap might create challenges... - >> which >>> we >>>> should understand upfront. I don't know all the foreseen use-cases >>> for the >>>> flow label, hence my earlier question on whether the question has >>> been taken >>>> to a larger audience (especially 6man). >>>> >>> I don't disagree, but the ip address used by the CPE is only used by >>> the CPE >>> sourced traffic, this is why I think it is safe to be used. One can >>> consider >>> all the IPv4 traffic of a household is a flow (for encap). In this >>> setup, >>> this is rather true and should not affect other flow label for native >>> v6 >>> traffic. >>> >>> I also want to hear more inputs. This is why I moved this discussion >> to >>> the >>> mailing list. We will also post this question to 6man. >>> >> ... FB: Great, and thanks for widening the audience. BTW/ - one could >> construct (theoretical) use cases, where the use of the flow label as >> CID would at least change network behavior, e.g. imagine that a provider >> would use ECMP and the hash would include the flow label (as initially >> anticipated when the flow label was introduced). With the assumption of >> SA and DA (Gateway and AFTR) being the same for all flows, the hash of >> the flow label would determine ECMP behavior - desirable or not, it >> would have an impact. So am curious to see whether there are real use >> cases out there. >> >> Thanks, Frank > > _______________________________________________ > Softwires mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/softwires > _______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
