mailing list archives
Beyond ITU and the Internet -- neutrality as to purpose
From: David Farber <dave () farber net>
Date: Sun, 14 Sep 2008 08:51:49 -0400
Begin forwarded message:
From: "Bob Frankston" <Bob19-0501 () bobf frankston com>
Date: September 14, 2008 1:13:45 AM EDT
To: <dave () farber net>, "'ip'" <ip () v2 listbox com>
Subject: Beyond ITU and the Internet -- neutrality as to purpose
We have two models:
· One in which you have an infrastructures own by service
providers so they can provide services. One example being a telephone
network. This is the classic telecom model.
· In the other you have a common infrastructure that is
neutral as to purpose. Anyone could provide services using this
commons. This is the Internet model.
The questions raised by Brett reflect the implicit assumptions that
are used to justify current policies. The most important point is that
providers seem to assume that they know what we want and design their
system to meet that need.
This is why it so very dangerous to leave them in the position to do
us favors. It used to seem absolutely necessarily for a telephone
company to provide us with circuits that had guaranteed performance
metrics. The Internet met none of those requirements. It gave users
the ability to fashion their own solutions.
Today we see Comcast engineering their network to satisfy the user
demands as they understand them. But they can’t understand them – all
they can do is give people what they say they want which is more of
what they had in the past (broadband) but the users can’t ask for what
they don’t have and those of us who do see beyond the consensus denied
the ability to move beyond Comcast’s imagination and to move beyond
what maximizes Comcast’s ROI as a service provider.
The important point is that there is no need for you to make any
assumptions about the devices in my house or the protocols I use.
I needn’t explain or any provider or protocol meister that I am
assigning identifiers to each of my thousands of photographs or that
I’m using 1024 bit GUIDs as identifiers. It’s up to me to resolve my
addresses to a path if I want to exchange bits.
In fact this what P2P is really about – reinventing end-to-end
connectivity despite what’s in the middle. Today’s P2P protocols tend
to be too tied to their applications though some like Skype have some
ideas we can learn from. I expect that there will be some that will
become standards like HTTP and HTML have. P2P is what I consider the
real Internet 2.
We need to have self-coined identifiers (like GUIDS) so that I can
define a relationship in isolation. They should be sufficiently unique
so that I treat them as being part of a large flat space but we don’t
have to assume there is a single naming space. The hard problem is the
authentication problem I cited with Caller-ID – how do you know who is
at the other end of a relationships and how to trust routing hints.
Given that these relationships are entirely independent of that path
and that I can act as if the addresses are globally unique we have the
kind of universal availability people assume. But we don’t really have
that now. My son (a very useful source of examples) found he couldn’t
use his home printer when using a licensed application which required
a VPN back into his school’s system (using his 32 bit laptop) for
authentication because the printer “name” is only the locally valid IP
address or zero-config name.
If my printer had a globally valid address and I chose to publish a
path then he could find it. Of course I would be able to control who
could find the printer since, by default, it is not universally visible.
We tend to confuse universal availability with the assurance that
there is a path even though many devices and networks are not always
connected and those inside local systems are hidden. We also
implicitly assume innumerability because today we have so few IP
addresses. And, because we started with the barn door open – systems
exposed by default – we tend to assume that search engines can find
These assumptions are similar to assuming we have a universal phone
book. But we don’t have one and, in fact, all attempts to do a
cellular phone book have failed because people like being in control
of their availability even if that control is limited to keeping the
name (phone number) secret.
We mustn’t confuse inarticulate expressions of what people say they
want with what is feasible. We see this again and again – they asked
for more CB radios and they got cell phones and few people would even
admit that really wanted CB radios.
But far more problematic is trusting providers with our future on the
assumption that they will give us what they want. They don’t even try
– they only give us what they think will maximize their ROI and
nothing at all more because that would violate their fiduciary
responsibility to their shareholders. We are customers not participants.
This why I keep emphasizing community connectivity – we can be our own
providers if only we weren’t naively accepting of those who insist on
keeping us dependent upon them.
From: Brett Glass [mailto:brett () lariat net]
Sent: Saturday, September 13, 2008 21:10
To: dave () farber net; ip
Cc: Bob19-0501 () bobf frankston com
Subject: Re: [IP] The ITU vs The Internet
At 03:41 AM 9/13/2008, Bob Frankston wrote:
>While traceability has many political implications ito?=s far more
problematic from a technical point of view. It means that one cano?=t
evolve protocols. Todayo?=s Internet compromises the end-to-end
principle by depending on an IP address from a central authority. Io?
=ve been arguing that we need to rediscover the Internet by assuring
that local networks are not dependent upon a central authority for
their names and addresses.
Are you suggesting, perhaps, a return to HOSTS.TXT?
(Note: for those who weren't involved in the Internet at the time, the
file HOSTS.TXT, circulated on the original ARPANet, was the way that
IP addresses were associated with host names prior to DNS. It quickly
became unwieldy as the network grew. DNS, at least, is decentralized
to the extent that the database is distributed and each authoritative
server has a sphere of authority.)
>I use the example of defining a relationship between a light switch
and a fixture in my house without any outside source of identity. This
relationship should still be meaningful if I take the switch with me
as a I travel around the world.
Why should it be? If you have a home control system in which the
number of local device addresses is, reasonably enough, limited to 256
or 512 or 1024, you shouldn't expect your device to turn on the lights
back home when it's connected in someone else's (though it may do
something if its address is meaningful to the network it's on).
>There cano?=t be a central registry of end-point identifiers nor a
single global network.
These are the things that users want. Users WANT universal
connectivity. And the "network neutrality" folks rave when one limits
connectivity even due to abuse of the network.
>I need to be able to use whatever transport is available with any
combination of logical and physical links. One big problem I found
with the current proocols is that if I am connected to two pipes I
cano?=t really make use of the power because each TCP connection is
limited to a single path.
It is actually relatively simple to multiplex connections among
multiple pipes. BGP allows it; so does multilink PPP (though with more
RSS Feed: https://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com
- Beyond ITU and the Internet -- neutrality as to purpose David Farber (Sep 14)