Bert, On Apr 15, 2010, at 11:12 MDT, Manfredi, Albert E wrote: > I meant, > > "In Brian's I-D 02, an edge router at the destination side would *NOT* be > able to tell whether the flow label had been set by a source host or an > intervening router." > > Bert > > -----Original Message----- > From: ipv6-boun...@ietf.org [mailto:ipv6-boun...@ietf.org] On Behalf Of > Manfredi, Albert E > Sent: Thursday, April 15, 2010 1:08 PM > To: Joel M. Halpern > Cc: ipv6@ietf.org > Subject: RE: Extracting the 5-tuple from IPv6 packets > >> -----Original Message----- >> From: ipv6-boun...@ietf.org [mailto:ipv6-boun...@ietf.org] On > >> If we can count on hosts setting the flow label with suitable >> granularity, then we can use the flow label (plus src and dest IPv6 >> address) in our ECMP and LAG hashes without having to look >> for protocol >> and port numbers. That avoids much complexity related to >> next headers >> and similar problems. And it is not subject to an attack by someone >> mis-setting the flow label field. >> >> The one obvious conclusion here is that if we want hosts to >> actually set >> flow labels, then we are largely preempting network modification of >> those flow labels. > > Looks to me like this is true as well. In Brian's I-D 02, an edge router at > the destination side would *not* be able to tell whether the flow label had > been set by a source host or an intervening router. So, that makes flow > labels unusable for end-to-end QoS. > > Question: if we don't want to specify different flow label ranges, e.g. to > show whether the lable was set by a host vs the ISP's network, then isn't > there a combination of flow label and traffic class that could accomplish > this? Something like this could be an option, instead of using the > traditional 5-tuples?
Or, better, (re-)acknowledge RFC 2474, a.k.a. DiffServ, is the sole go forward mechanism for CoS treatment, meaning PHB's/treatments are to be based only on the IP Traffic Class field. The industry seems to have at least implicitly acknowledged that more granular, IntServ-type signaling (perhaps, the original intent of the flow-label?) aren't scalable in the IPv4 Internet and, likely, IPv6 Internet as well. Thus, we should move on to a more appropriate use of a IPv6 flow-label? -shane -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
