!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()
------------------------------------------------------------
SMART Letter #3 -- February 15, 1998
For Friends and Enemies of the Stupid Network
Copyright 1998 by David S. Isenberg
This document may be redistributed provided that
the 11 lines containing this notice accompany it.
isen@isen.com -- http://www.isen.com/ -- 1-888-isen-com
It takes SMART people to design a Stupid Network
------------------------------------------------------------
!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()
SMART People,
This swirl of SMART ideas is one of the biggest
highs of my life -- and I've taken some pretty strong stuff!
But I am still struggling to find a good way to realize
the value of (and collectivize) the discussion.
On one extreme, we could just have an email reflector, and
everybody could see everybody's every comment. We tried this,
and we became victims of our own success. Our mailboxes filled
up, and many otherwise interested people soon found it too
much hassle to keep up. It became "so popular that people didn't
want to go there." Then we tried posting to an archive site.
This caused traffic to drop off, and the discussion to become
less vital. It was too hidden. And I found that once I
had fired up my 'scape 'splorer, there were lots of other
places to go . . . So I'm going to try something new.
Here's my first attempt at a "moderated list." Boy it is
a LOT of work! I don't know if I can keep doing it.
Also, I don't entirely trust my own editing -- I'd feel
more comfortable if you all could read my edited comments,
and have the whole thing available, too!
I note that there are two wonderful, articulate new comments
on the archive site -- http://www.isen.com/SMARTlist/ -- by
Kai Matthews and Brian Gygi. I have excerpted them below,
but I feel much better that you can go read the originals
as their authors posted them.
So please let me know what you think of the below. I'd
like it a lot if you'd post your answers to
smart-discussion@isen.com -- you can post it to me, too,
but if you think it might have general interest,
consider posting to smart-discussion as well.
Thanks,
David I
. . . . . . . . . .
From: Cameron Simpson
I'm a SysAdmin of the UNIXish persuasion . . . I've a
loathing of complex systems . . . [I see] UNIX and TCP [as]
essentially simple systems with tools layers on top . . .
[A] downside of complex things (aside from the difficulties
involved in ensuring a complex spec is actually consistent and
complete) is that it readily lends itself to becoming proprietry or
monopolised. Some specs are so complex that the cost of bootstrapping
to basic compliance is often beyond the individual or small company
seeking to implement a tool in the area. Nasty.
. . . . . . . . . .
From: Neal Goldsmith
Your point about stupid networks is intelligent. Trying to make the system
more intelligent by adding complicated functionality to the network itself,
is like trying to make a brain more intelligent by making neurons more
complex (and thus into bottlenecks). Rather, it is simple, unencumbered,
consistent, rapid flow that facilitates the emergence of a complex
intelligence out of a brain. Like Kelly's Out Of Control -- enable it and
then get out of the way!
. . . . . . . . . .
Bravo! . . . your four principles cut to the quick. Very nice.
They ring with the same authority as Kevin Kelly's "Nine Laws of God,"
and they are less than half as numerous.
. . . . . . . . . .
From: "Chuck Gritton"
This is a fascinating thesis. A question though for you - using
your automobile analogy, how does one configure your car to ensure
a no-crash, fixed-time arrival at the destination? I would think
there are some business applications that would want that level
of service and be willing to pay for it.
[Chuck is the CTO of Coherent Systems. I sent him some answers to his
question (but only to suck up so he'll hire me for pro-sulting :-)).
Do other SMART people have additional answers for him? . . . David I]
. . . . . . . . . .
From: Leslie Good
[The Stupid idea] resonates because we are all suffering from the those
who want and need to control. We have to keep the Internet free! I worry
that if we are not vigilant, this will become a still birth, stifled by the
likes of Microsoft . . . It seems to me to be a moral issue. Encrypt all
your banal email just for the CIA to fume about.
[Seems that Good is an enemy of "perfect" now :-) . . . David I]
. . . . . . . . . .
From: Christopher Locke
. . . congrats on having escaped the evil klutchez of Mamma Belle.
I just put up a page about this new book I got included in called
"The 21st Century Intranet" by Jennifer Stone Gonzalez
(see http://www.rageboy.com/21st.html )
I would love for you to pass this along to the list . . .
[Here is a fabulous "scenaric" quote from the Gonzalez book:
"Quantitative analysis works wonders when all the assumptions are
accurate and the variables selected are the right ones. This happens
approximately once every million years."
Chris Locke is the alter ego of the infamous Rage Boy -- www.rageboy.com --
and the author of "Entropy Gradient Reversal" the satirical e-rag that
occasionally makes me laugh so hard I have to change my underwear.
David I]
. . . . . . . . . .
From: Brian Gygi
Subject: Stupidity in World History
What David Isenberg refers to, catchingly, as 'stupidity' merely one end
of the continuum from simplicity to complexity that has been a part of
almost dynamic systems since physical laws were invented. There is a
tendency among scientists, the Santa Fe folks pre-eminent among them, to
regard more complex systems are somehow 'better', ignoring the fact that
there is an eternal trade-off between efficiency of a system and the cost
of implementing that efficiency, and of course the ability to respond to
change in the driving forces.
Now of course the question for the Internet, is what are those driving
forces? Unfortunately, it seems that economics is the driving force,
which means that we are going to see precisely the kind of changes Mr.
Isenberg doesn't want.
[Brian, eloquently put! My idea is that there is a new economics
afoot, but you are right that the complexifiers, the incumbent interests,
have lots of motivation to introduce complexity that takes control out
of the hands of the end users. They do it in the name of making the
Internet safe for children, or in the name of Quality of Service, or mission
critical applications, or national security, or . . . but as you astutely
observe, economics is the driving force. And I do not know another era
when an economic good has been getting so cheap or so accessible so
rapidly. Do you? Conceivably, it could be different this time
. . . David I]
. . . . . . . . . .
From: Gordon Cook
The stupid solution is best effort delivery where you keep throwing bits
at someone in the hope that they will eventually get there. the
providers hate this because it is a commodity based service with no
profit.... study the ANX [Bellcore's Automotive Network Exchange,
an actual attempt at an "industrial strength IP network] .....see why
mission critical stuff doesn't like best effort.....also over coming
the latency involved in IP telephony requires some QoS to guarantee
packet delivery.
beware of chopping everything to fit your single procrustean bed. get a
book on the tcp/ip protocol..... if I am right, this protocol is
exceedingly intelligent. but also exceedingly flexible where atm is not.
better learn more about best effort delivery and the business models that
can be supported by it before you bash QoS.
beware of being too dogmatic..... :-)
[Gordon Cook is the force behind The Cook Report, the ultimate
Internet insider's newsletter. If you have to have the
latest NANOG natterings, or the hidden skinny on Universal Service
initiatives, or a persnickety perspective on peering, you'd be advised to
subscribe. And he liked "Rise of the Stupid Network," so he
clearly knows what he's talking about! . . . David I]
. . . . . . . . . .
From: Jock Gill
[The bed that Jock Gill wants to put Stupid into is anything but
Procrustean. Jock believes that Stupid concepts (such as decentralization
and control at the edges) can be applied to the electric power grid,
and now, to financial networks. Is he stretching the occupant to fit
the bed, or are these ideas broader than I originally
drew them? . . . David I]
. . . [Current practice is that l]arge industrial loans [are made for
centralized facilities, because] the rational metric says the costs
are lower and the efficiencies are higher. Fund a large coal fired
plant, yes, but [its] impossible to loan the same amount to
1000s of individuals to finance distributed micro power
projects supporting micro enterprise and micro e-commerce.
Isn't distributed micro solutions the same as putting the
smarts at the edges? The closest example I see in a western economy
is the American way of financing private home ownership.
This leads to a more general question: is concentration in the
centers a law of nature which leads to people adopting smart
networks? Are stupid networks counter intuitive to this
'clustering'? Or are the metrics wrong so the 'centrist' world view
only 'looks' rational.
Which is to ask can we extend the Stupid Network idea beyond the
limits of purely physical networks to other sorts of networks:
political? financial? social? Do we have to integrate an and/both
reality: a mix of centralized smarts with edge smarts?
I expect we do.
. . . . . . . . . . .
From: Douglass Carmichael"
[Doug speaks to some of Jock's issues . . . David I]
I've recently read a book, The Collapse of
Complex Societies, Joseph Tainter, Oxford University press, who argues that
the marginal utility of complexity diminishes rapidly, and that elites chose
complexities as social solutions because the payoff is so high in wealth
transfer. This seems to argue for the real utility of simplicity.
David Hackett Fischer's the Great Wave argues that policies lead to
inflation which breaks he back of the social system, and that there have
been three major breaks, 1400, 1600, 1800 and that we are at an all time
inflationary high, with lots of predicted turbulence. this also supports the
need for simplicity.
the [Year 2000 Problem], now getting really thick and serious with the FAA
announcement yesterday and Clinton's appointment of John Koskinen, indicate
that we may have extended ourselves with computation into a domain of
fragility without any recognition of where we were getting ourselves.
. . . . . . . . . .
From: Dan_Tebbutt@acp.com.au
I disagree about . . . your assumption that QoS is necessarily a matter
of centralised control . . . QoS is inherently too complex to be
implemented in a global 'one size fits all' manner. It will only
ever be in the most loose form across the network . . . if it's
benefits are really as presumed it will prove popular.
[Y]our argument against optimising IP networks contradicts your
leading statements about the advances we've had from optimising
consumption of gasoline.
[I am delighted that Dan has the courage to disagree, and it is
intelligent, information-bearing disagreement. And he did get it
right about the complexity of QoS! But Dan, I did not say that
QoS *was* a matter of central control, only
that I was suspicious that it was moving in that direction.
You may well be right about its impending
lack of universal implementation.
Also, the contradiction between Stupid Network and Automobile
logic was deliberate -- more a juxtaposition than a contradiction.
Perhaps I should have motivated it more clearly . . . David I]
. . . . . . . . . . .
From: Gary Hughes-Fenchel
Consider: I own the wires running to your telephone, and I own
the CO. . . . Keeping the network simple through your principles
may make good sense from a societial standpoint, but complicating
the network - and thereby forclosing competition - may make more
sense to your service provider.
. . . from the perspective of a product company . . . I think the
only [issue would be] how many potential purchasers there were for
their equipment. Let's say you need one box X to provide some new
feature. Let's say the market demand for the service peaks at
500,000 units at some price range. A product provider
cares a great deal whether the 500,000 units are going to be
purchased by 500,000 purchasers, 1,000 purchasers, or 10 purchasers. . . .
[Gary asks that I tell you that he speaks for himself, not Lucent
. . . David I]
. . . . . . . . . .
From: "The Adam N. Rosenberg"
I'm no expert on these sorts of networks as my last involvement
was the Wideband Packet Abortion back at AT&T Bell Laboratories . . .
But the Ethernet concept rubs me the wrong way, the notion
that we should build a network one hundred
times bigger than the traffic and not worry about any kind of smart
signaling protocol. If we want ten times as many bits, then we'll
build a network ten more times bigger than the solution we have now.
In software, that gives us a text editor that won't run on a
machine with four megabytes of memory, Microsoft Word . . .
I think that each Word user should have to wire 56 623 104 little
ferrite cores (that's six megabytes plus parity bits) to find out just
how inefficient his tool of choice really is. [Maybe we should
put Bill Gates and Andy Grove to this task :-) . . . David I]
Sure, I'm somewhat more cavalier about memory than I was when I
had 64 kilobytes, but I think through my big memory arrays all the
more carefully. I find that I can do more than twice as much with
twelve megabytes as six in a big simulation if I do proper planning.
Does that mean you're wrong? Probably not. But it is a point of
your philosophy that I believe isn't "universally" right. Of course,
you didn't say it was right everywhere and I doubt you want to go back
to getting ten miles to the gallon, either. But there is a tradeoff
that has two sides and most designers only see one of the sides.
. . . . . . . . . .
From: Kai Matthews (kaimuse@apocalypse.org)
[Stupid is] an excellent addition to the subject of User-Centered Design
(as elucidated by Landauer, Kreitzberg, Kapor, Dertouzos, etc.)
I'm not a network pro or even a programmer, but then most of the folks who
use and increasingly need the Net aren't either, and I think we have some
right to be heard, and our own kinds of intelligence, our own perspectives,
to contribute on this subject.
. . . Well, there's actual simplicity, and then there's *apparent*
simplicity...
Both desirable goals. Actual simplicity is of course intimately related to
underspecification...
I think *apparent* simplicity, on the other hand, can often be achieved by
sufficient translation and/or encapsulation of particular skillsets into
other skillsets - the complexity is still there "under the hood", but the
users encounter it in a form more-or-less native to their knowledge-base,
their skillset(s), and therefore accessible to their control. Good
interface design is one, but not the only, aspect of this accessibility.
. . . I'm always glad to discover people thinking along the lines of
simplicity and transparency. This isn't an optional, "it'd be nice if we have
the time" consideration; rather, it's essential to the evolving inclusiveness,
usefulness and usability of the Net.
[Kai's long, articulate, thought-provoking musings on Stupid are on
http://www.isen.com/SMARTlist/0361.html . . . David I]
. . . . . . . . . .
From: Steve Norby
[The below is excerpted from a letter to Bob Lucky about his article,
"When is Dumb Smart?" Steve says that these words represent personal
perspectives, not US West's. A SMART reader will not confuse the
below for official telco anything . . . David I]
I grew up in the great days of telephone networks when we
started adding common control processors and digital carrier in the
"plant". . . . our revenues were essentially based upon how much
we spent . . . Rewards were provided for being economically dumb.
I particularly found it interesting that a local subscriber had to pay
more for "touch-tone" dialing service when it actually cost more
to provide circular pulse dialing through a modern switch.
When the state utility commissions began to scrutinize our industry
closer, we found ways of relabeling investments as long-distance
oriented so that the game would shift to the toll settlements process.
. . . Of course, once competitive long distance services
became fashionable, local telephone costs appeared to be rechanneled
into interexchange access investments so that we could ensure that the
revenue kept rolling in.
[I] evolved into a garden variety central network intelligence control
bigot -- a la AIN. I felt secure in my AIN-like perspective until the
day in 1991 that I went to visit a prospective customer at Washington
University that was doing work in teleradiology over cell relay transport.
When I saw how his service control architecture was designed around
intelligent end stations and somewhat unintelligent network transport,
I was enlightened into viewing the network in a new way.
. . . The intelligent user/dumb network concept stewed in the backburner
of my brain as an interesting but unfounded perspective for a couple more
years. I didn't give it a whole lot of thought until the day I was
shown a demonstration of the World Wide Web from a coworker of mine back
in 1994. At that point, I realized that the carrier network view of
intelligence and control was doomed. I looked at the Internet in a
whole new way as a real conveyance for telecommunications.
Needless to say, I am now a true disciple of the not-so intelligent
network cult -- I think the term stupid is a little too harsh -- where
the end users will possess the real service control. While the
empowerment of the masses may be conveyed best through Trojan horses
such as Windows 95/NT/CE/98, JAVA and other enablers requiring
intelligent user terminal equipment, it is inevitable. Intelligence is
moving back to the periphery of today's network.
*--------------------isen.com----------------------*
David S. Isenberg isen@isen.com
d/b/a isen.com http://www.isen.com/
18 South Wickom Drive 888-isen-com (anytime)
Westfield NJ 07090 USA 908-875-0772 (direct line)
908-654-0772 (home)
*--------------------isen.com----------------------*
-- Technology Analysis and Strategy --
Rethinking the value of networks
in an era of abundant infrastructure.
*--------------------isen.com----------------------*
Date last modified: 16 Feb 98