Re: I-D Action: draft-han-6man-in-band-signaling-for-transport-qos-00.txt

[Tom Herbert <[e-mail]>]

On Thu, Oct 19, 2017 at 3:09 PM, Toerless Eckert <[e-mail]> wrote:

> On Thu, Oct 19, 2017 at 03:00:35PM -0700, Tom Herbert wrote:
> > Bert, I don't follow. One of the big wins using the 3-tuple is that it
> can
> > still indicate a flow even if all the transport transport headers and
> > payload are encrypted. So this should allow specifying QoS without
> > revealing anything about the content except that the sender wants to
> group
> > certain packets together as a flow and apply QoS on them.
>
> Problem with flow label is same as with hop-by-hop option: burned because
> nobody should dare use it for something useful with all the inconsistent
> code out in the field.  Nothing against redefining it (with new encoding)
> with a precise set of MUST/MUST-NOTs once we agree on some way to use it.
>
> Toerless,

I know that EH is an issue because some implementations drop packets having
them. But what is the problem with flow labels? Non-zero flow labels are
being used everyday on the Internet without problem, and I know of at least
one big DC deployment that is productively using flow labels for ECMP in
their network.

Tom