ietf
[Top] [All Lists]

Re: Naive question on multiple TCP/IP channels and please dont start a uS NN debate here unless you really want to.

2015-02-06 13:52:04

On 2/6/15, 2:27 PM, "Brian E Carpenter" 
<brian(_dot_)e(_dot_)carpenter(_at_)gmail(_dot_)com>
wrote:

On 07/02/2015 08:05, Piers O'Hanlon wrote:

On 6 Feb 2015, at 18:24, Richard Shockey wrote:


Fine now how do you get the labeling/queueing across the AS boundary?
I
don¹t know any ISP that accepts or recognizes the packet labeling of
another AS.

Sure - that's another whole ballgame! A number of ISPs blow away the
DSCP bits in packets from and to the home, as I understand they use
their own set of DSCPs internally.

That is entirely in keeping with the diffserv architecture, which is
explicit that DSCPs are domain-specific and that traffic may be
reclassified at domain boundaries. (Which is what operators wanted
when diffserv was designed.)

But agreements of use across boundaries aren't that clear and probably
wouldn't generally be extended to end users.

Agreements across boundaries require mutual trust, so it's to be
expected that ISPs will reclassify traffic arriving from subscribers.
For ISP/ISP boundaries, see
http://tools.ietf.org/html/draft-ietf-tsvwg-diffserv-intercon

I guess they're also using things like MPLS, or SDN (e.g. Google B4)
for traffic engineering.

Diffserv isn't traffic engineering, however.


Well Brian after Feb 26 we don¹t know what DIFFSERV will be.  It may be
illegal unless you can define what a ³specialized service², ³reasonable
network management² or ³commercially reasonable²  actually is.



<Prev in Thread] Current Thread [Next in Thread>