RFC1883 and ipv6 spec v2

[A.N.Kuznetsov]

Hello!

> 1/ expensive to do - gotta stick the old flow ID somewhere in the meantime

Right. Hence, this practice will be automatically deprecated :-)

If flow labels will loose their end-to-end meaning at least
sometimes, they will be useless for rsvp. Is it correct? 

> 2/ the diff serv work (some of which is already being tested in the 
> field) may depend of the changability of some of the packet header at
> cloud (e.g. ISP) borders

I was not aware about diff serv work, but Cisco tag switching
can be made efficiently or by link layer (ATM) or by inserting
tag between ll header and network header (that is made for IPv4 in nay case).
Mangling flow label is not necessary at all.

Alexey Kuznetsov