Re: [BEHAVE] IPv6 Flow Label

[Hesham Soliman <hesham@elevatemobile.com>]

On 6/05/08 6:38 PM, "Rémi Després" <remi.despres@free.fr> wrote:

> Brian E Carpenter wrote :


> So the rule is really only that the flow label
>
> must be delivered *exactly* as it was set by the source host (and the
>
> default setting is zero).
Right.
However, between source and destination hosts
> *of IPv6 packets*.
On an A-TURN-B path, the *destination host* of a packet
> from A is TURN
(not B).

> TURN has to obey that rule like any other
> device
> that forwards packets.
TURN is in this respect like any other function that
> operates with
different 5-tuples on its upstream and donwstream interfaces
> (above the
IP layer).




  Following the discusion, this is IMHO the right
> logic:

TURN nodes may freely support or not support RFC 3697 flow labels, as
>
sources on their outgoing interfaces,.
Independently, they may freely ignore
> received flow labels, or process
them to optimize their transport flow
> recognition, on thir incoming
interfaces.

If RFC 3697 flow labels are
> implemented by the two end nodes and by an
intermediate TURN node, then
> packet classification can take advantage of
flow labels in all traversed
> routers, be it with changed flow label
values across the TURN
> node.




=> I don't understand two points:
- Why is the IPv6 address/header being re-written?
- If there is a good answer for the above question then why would you ignore
the flow label and not copy it across?

It is one thing to terminate the connection and restart it with the remote
node, but I don't see any reason for re-writing a header _and_ ignoring the
original content of that header.

Hesham



Rémi



_______________________________________________
Behave mailing
> list
Behave@ietf.org
https://www.ietf.org/mailman/listinfo/behave