more on Coalition Asks FCC to Ensure End-to-End

[Dave Farber <dave () farber net>]

------ Forwarded Message
From: Bob Frankston <rmfxixB () bobf Frankston com>
Date: Sat, 14 Dec 2002 22:34:10 -0500
To: dave () farber net, "'ip'" <ip () v2 listbox com>
Cc: Karl Auerbach <karl () cavebear com>, Glenn Fleishman <glenn () glennf com>,
David Weinberger <self () evident com>, David Reed <dpreed () reed com>
Subject: RE: [IP] more on Coalition Asks FCC to Ensure End-to-End

Ah, QoS -- Karl is right in pointing out the slow first (please, first
mile, we are not just there to be delivered to) mile is a
rationalization for QoS. I agree with Karl's conclusion that we'll see
lots of resistance to E2E. Asking for QoS is an example of just this
kind of resistance so it is worth being explicit about how and why.

I've been doing a lot of writing on the topic but mainly in small
discussions where I can be free to use a ranting style. Better to read
Glenn's coherent comments in
http://blog.glennf.com/gmblog/archives/00000285.htm and I know that
David Weinberger is also working on comments.

Just as there is a danger that a call for openness can result in more
regulation rather than transparency, QoS is part of the crypto-bellhead
attempt to turn the Internet back into the PSTN with its oh-so-special
treatment of voice streams.

The proper response to current first mile limits is, as I've pointed
out, to recognize that it the problem is due to intransigence and
conflict of interest. There is no technical reason for such limits.
Sympathizing with those holding connectivity hostage (also known as the
Stockholm Syndrome) seems responsible but is self-defeating.

QoS gives the middle the power to decide what traffic is more important
than other traffic and it presumes that either one can examine the
packets to determine whether they are VoIP packets or that the ends can
flag which packets are important. The former is problematic because we
are at the start of learning about VoIP and the initial implementations
(mostly by Cisco as with ones I got from http://www.vonage.com) are only
experiments and there is lots of room for other approaches. Thus one
can't identify the VoIP traffic without explicitly defining it as
neoPSTN. Of course, responsible users should encrypt all their traffic
and those connecting to offices probably use VPNs anyway. Even if the
users are trusted to mark their own important packets (all of mine are
important!) the VPNs will hide that information.

The good news is that QoS can't work even at the edges. And the better
news is that there is no problem at the edges that can't be solved by
adding some electronics on the wire. Of course, as long as effort is
going into QoS we are getting no new capacity and just more arbitrary
discrimination.

QoS, like "Broadband", only provides cover for the incumbents to stave
off transparency.

Again, the very good news is that if we have providers (as I pointed out
in my previous message) whose incentive it to provide incrementally more
capacity we have the virtuous cycle of Moore's law (or what I call "just
let me buy a little more at any point rather than having the Hobson's
choice of 1.5mbps or 768kbps") and we have a decade of pent up
technology ready to be deployed. As we've seen, latency melts away as
capacity increases. At least as long as we don't to spend capacity
trying to determining which packets require special treatment and
buffering the traffic to create queues for sorting traffic.

Even with minimal DSL assumptions and only a few mbps on the copper path
(again, there is no distance limit because regenerating the signal is
easy) there is no problem with an 8kbps voice stream. But why not demand
a pair of 64KBps streams or 6 channel audio?

I've been accused of being idealist but I am a pragmatist since I don't
believe anything complicated can work. We are taught that there are no
simple solutions to complex problems yet there is nothing simpler than
the Internet.

Asking for more special treatment be it QoS, Broadband or Faux-Openness
is seductive but it cedes the future to the incumbents. Instead we must
focus on simplicity in the captive segment at the first mile. {See my
previous message for separation comments}

PS: Is First Mile (or even Last) an acceptable idiom or do should I say
Kilo?

[ I have been, ever since I went to the FCC, using the term first mile
(kilo) to distinguish the problem from the traditional ilec attitude -- we
deliver to you what we want (the LAST mile) as opposed to the internet
attitude and the users one -- we supply and get -- the first mile djf]

Bob Frankston
http://www.Frankston.com

------ End of Forwarded Message

-------------------------------------
You are subscribed as interesting-people () lists elistx com
To unsubscribe or update your address, click
  http://v2.listbox.com/member/?listname=ip

Archives at: http://www.interesting-people.org/archives/interesting-people/