On Fri, May 31, 2013 at 11:02 PM, bingxu...@gmail.com <bingxu...@gmail.com>wrote:
> [Qiong] I'm trying to answer this question from operator's aspect. When we > are trying to encode semantics e.g service type, subscriber type, etc., in > some flag, it should have the following features: > 1) It should exist and be encoded from the beginning. Otherwise, we still > need to introduce dpi-related system to identify certain services, which I > believe is not scalable. > 2) It cannot be changed along the path. Otherwise, there will be trust > issue and may be changed to a different value without notice. > 3) It can be easily treated in existing routers. In the ideal way, it does > not need to introduce new functionality so that all our existing routers > can deal with it. This can greatly reduce our upgrading cost. > These are major reasons that we embed semantics in prefixes. But I'm > still open suggestions if there is a better choice. > DSCP does all of those today. All you need to do is ensure that you always mark the packets on ingress.
-------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------