Re: My comments to the press about RFC 2474

Richard,

This will be my last message on these points, which were beaten to death
in the diffserv WG some years ago.

assured or expedited services, except nobody really knows how that might 
actually work in a real scenario (or maybe they do, and it's just us humble 
developers who don't.)


I can't speak for those who have implemented EF and AF queuing algorithms;
they will have to speak for themselves.

Am I missing something when I find a gap in the dialog?


I don't see the gap. There are a few references to possible differential
payments in some of the informational documents concerning diffserv, and nobody
denies that differential pricing is possible. There are no elements in the
normative specifications of diffserv that serve in any way to support
accounting, pricing or charging. We can't control what choices the carriers
in one particular country such as the USA adopt; and it's not our business.

The fact that journalists don't read the bit at the beginning of each
RFC that defines its status is something that we've been used to for
many years.

Regards
   Brian

On 2010-09-04 10:36, Richard Bennett wrote:

 Let's go back to your original comment, the one that Russ has quoted
elsewhere. You said: "It has been consistently hard to explain that
diffserv is not a prioritisation scheme, even within the technical
community, let alone to the regulators and the media." Your
clarification is that "DiffServ deals with multiple queuing disciplines,
which may or may not be priority based" and you get into EF and AF that
in most implementations will end up using dedicated facilities of some
sort, although service providers may be able to use these dedicated
facilities for generic traffic if there's no EF or AF to send.

I can see why it's hard to explain this subtle distinction. You're
saying here on the list that DiffServ is a higher level abstraction than
prioritization that neither requires nor precludes prioritization to
implement premium services. That's fine, but if you want to say that
DiffServ is *completely removed from the realm of prioritization,* you
have to explain away the grandfathering of IP Precedence into the
standard. I think that you mean to say that DiffServ is *more than* a
prioritization scheme, it's a general architecture for Internet QoS."

But even that's not correct. In fact, DiffServ is an Internet QoS
architecture that explicitly offers priority-based services by design,
and may also offer other types of assured or expedited services, except
nobody really knows how that might actually work in a real scenario (or
maybe they do, and it's just us humble developers who don't.)

Russ said to the press that he considers AT&T's belief that the RFCs
envisioned payment for premium services implemented over DiffServ or
MPLS to be "invalid." I find this puzzling as there are numerous
references to payment for premium services in the RFCs AT&T cites, such
as RFC 2638:

"At the risk of belaboring an analogy, we are motivated to provide
services tiers in somewhat the same fashion as the airlines do with
first class, business class and coach class. The latter also has
tiering built in due to the various restrictions put on the purchase.
A part of the analogy we want to stress is that best effort traffic,
like coach class seats on an airplane, is still expected to make up
the bulk of internet traffic. Business and first class carry a small
number of passengers, but are quite important to the economics of the
airline industry. The various economic forces and realities combine
to dictate the relative allocation of the seats and to try to fill
the airplane. We don't expect that differentiated services will
comprise all the traffic on the internet, but we do expect that new
services will lead to a healthy economic and service environment."


Am I missing something when I find a gap in the dialog?

RB

On 9/3/2010 3:11 PM, Brian E Carpenter wrote:

Er, exactly what in your quotation is incompatible with what
I wrote:

Diffserv deals with multiple different queuing disiplines, which
may or may not be priority based.

?

Regards
    Brian Carpenter

On 2010-09-04 09:34, Richard Bennett wrote:

   Brian's paper on DiffServ confirms the fact that prioritization is
part of the
standard. Here are the two relevant quotes:

"In the original design of IP [33], a byte known as the “type of
service (TOS)
octet” was reserved in the header of every packet. This was defined
to contain
two important fields: a three-bit “precedence” value and three TOS
bits. The
precedence was intended as a simple priority marker, where priority 0
got the
worst treatment and priority 7 got the best." (p. 1480)

"The Diffserv working group has taken the approach that a few
fundamental PHBs
should be standardized early. These should derive from some existing
experience
(primarily from limited deployment and experimentation with use of
the IP
precedence field to select forwarding behaviors) and might be
implemented using
a variety of specific mechanisms. The PHBs standardized so far are as
follows...

"• Class selector behaviors: here seven DSCP values run from 001 000
to 111 000
and are specified to select up to seven behaviors, each of which has
a higher
probability of timely forwarding than its predecessor. *Experts will
note that
the default behavior plus the class selectors exactly mirror the
original eight
IP Precedence values.*" (p. 1487)

This is very straightforward.

RB

On 9/3/2010 1:06 PM, Brian E Carpenter wrote:

Richard,

Diffserv deals with multiple different queuing disiplines, which may
or may not be priority based. Please read RFC 2475 and if
you like, B.E. Carpenter and K. Nichols, Differentiated Services in
the Internet, Proc. IEEE, 90 (9) (2002) 1479-1494.

     Brian

On 2010-09-04 07:57, Richard Bennett wrote:

  DiffServ is a prioritization scheme, Brian, how can you say it's
not?
IntServ is a reservation scheme, and DiffServ attempts to provide
desired PHBs in practice by sorting packets into priority queues and
invoking appropriate Link Layer  facilities, which are in most cases
priority-based, such as 802.11e traffic classes.

What on earth could the value of DSCPs be if they didn't map to
traffic
classes in the data link?

RB

  Brian E Carpenter<brian(_dot_)e(_dot_)carpenter(_at_)gmail(_dot_)com>   
wrote:

Russ,
It has been consistently hard to explain that diffserv is not a
prioritisation scheme, even within the technical community, let
alone to the regulators and the media. I think your comments as
quoted are as good as we can expect from journalists.
It should be a matter of concern to all of us here that the US FCC
isn't confused into regulating the technology. It would set a bad
precedent for regulators in other countries. I am making no comment
as to whether they should regulate carrier's charging practices;
that's
entirely a national matter that shouldn't concern the IETF in any
way.
Regards
Brian Carpenter
On 2010-09-03 05:47, Russ Housley wrote:

I want the whole community to be aware of the comments that I
made to
the press yesterday. Clearly, these comments do not represent IETF
consensus in any way. They are my opinion, and the reporter was
told to
express them as my opinion.

One thing that I said was not captured quite right. The article
says:
"With services that require certain speeds to operate smoothly,
such as
Internet telephony, calls are given precedence over TV, Housley
said."
I actually said that DiffServ can be used to make sure that traffic
associated with applications that require timely delivery, like
voice
and video, to give preference over traffic associated with
applications
without those demands, like email.

The whole article is copied below, and it is online here:
http://www.nationaljournal.com/njonline/tc_20100902_7144.php

Russ

-- 
Richard Bennett
Senior Research Fellow
Information Technology and Innovation Foundation
Washington, DC


_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf