SMART Letter #28
October 8, 1999



            SMART Letter #28- October 8, 1999

             Copyright 1999 by David S. Isenberg -- "making problems for 'solutions providers'" -- -- 1-888-isen-com




> A User-Owned, Self-Organizing Network

> Why Carriers Want the Network to Know Content Type

> Conferences on my Calendar, Copyright Notice, Administrivia


ROOFER MADNESS: Rooftop's wireless technology bypasses local telcos, and

creates a user-owned, self-organizing, service-providerless network.

By David Isenberg

Nokia's early September move to acquire Rooftop Communications may prove

brilliant - if Nokia can manage radically disruptive technology. Regardless,

Rooftop has convincingly demonstrated that the role of communications

service provider is waning, and that self-configuring, self-service networks

are close at hand.

Rooftop has produced a spread-spectrum radio plus packet router cleverly

described by its trademark, "Unplug and Play." This device gives Internet

service providers (ISPs) a link to customers that is not mediated by any

infrastructure provider. In principle, the DSL-speed network formed by

multiple devices could even run without an ISP, or with multiple ISPs -

indeed, multiple ISPs would make such a network more robust.

Network architecture emerges as a multihop, multipoint network. Lauren Hipp,

Rooftop's EVP of marketing, calls it a "peer-to-peer distributed networking

scheme, a routed mesh." Each device is both a customer's access device and a

network infrastructure node. Adjacent customers (up to a couple of miles

apart) form a skein that is a natural extension of the Internet.


In theory, customers will be able to buy a device, take it home, plug it in,

and log onto the Internet. In practice, today's installation issues include

geography, roof rights, and how good a customer is with ladder, drill and


Each customer node needs a line of sight to at least one other node, and

preferably to more than one. Unlike LMDS, a line of sight directly to a

point of presence is not necessary - this is a benefit of multihop

architecture. Even so, there are antenna placement issues; in the best case,

the antenna can be placed in a window from which it can 'see' another node -

but, more typically, rooftop is the name of the game. For longer distance

hops, a directional antenna can be used.

While suburbs with one-story houses and many tall trees might never support

a Rooftop network, Phoenix and Los Angeles are ideal environments. The San

Francisco Bay Area already supports several beta nets. The world's

developing nations often provide ideal environments - there are pilot

networks in Morocco and Guinea. "It is critical technology where there is no

existing infrastructure," Hipp says.

Adding customers improves throughput and robustness; a dense net has many

alternate routes from a customer's premises to the ISP's 'airhead.' But slow

acceptance is a palpable risk.

"The technology is promising, and looks scalable," says Bob Lucky, vice

president of applied research at Telcordia. "But it could never get started

because you need customers to create the mesh. I'd be afraid that one day

the critical person between me and my ISP would say, 'Sorry, I just got DSL,

so I'm going off the air.'"

Hipp rebuts this thinking. "We have a lot of demand," she says, "and a

jump-start promotion for interested ISPs. We're aggressive about reaching

the consumer marketplace. We have two priorities: cost and performance."


Today, Rooftop has two devices priced for small business: a $2,000 unit and

a more sophisticated and powerful $5,000 device. Both, says Hipp, are on

steep cost-reduction trajectories. They're being redesigned for better

integration. Increasing volume will help, and components are getting

cheaper. "Radio technology is jumping onto the digital price curve," she


There's room for improved performance, too. Each hop adds some 50 ms of

latency. This limits the practical number of hops to three and it severely

limits the quality of Internet telephony. But, Hipp says, "In the future, we

think we can support telephony-quality voice. We haven't tuned IROS

[Rooftop's proprietary Internet Radio Operating System] to optimize latency


John Giannandrea of, a Silicon Valley ISP, is an early adapter of

and testbed for Rooftop technology. He says that it gives new meaning to

community networking. "The network grows organically," he says. "The ISP can

get out of the business of planning network expansion."'s wireless customers include a gourmet restaurant, Global Village

Cafe, whose owner (a former techie) wanted little to do with the incumbent

telco. Other customers find a Rooftop installation easier than obtaining

DSL - and DSL is more readily available in California than anywhere. Another

factor is that Rooftop input/output is inherently symmetrical.

Will Rooftop pose an "Innovator's Dilemma" for Nokia? The "wireless DSL"

aspect may be a bit disruptive. But the radical disruption will come as it

fulfills its potential to make local telcos obsolete. Will Nokia be able to

capitalize on this, or will it "make sense" for it to limit or kill the

product because it threatens its current customers, the world's telcos?

The article above appeared in the October 1, 1999 issue of America's

Network. Copyright 1999 Advanstar Communications.



by David S. Isenberg

Nicholas Negroponte, head of MIT's famed Media Lab, in a recent contribution

to the discussion board, wrote that data networks "have no

way of knowing which bits are voice and which are something else."  Yet

Negroponte once wrote (Wired 6.06, June 1998) about how some bits (e.g.,

your pacemaker reading when you are having a heart attack) are infinitely

more valuable than other bits (e.g., the twinkly background of a banner ad).

Will telcos charge according to the cost of sending bits or by the value of

those bits to the customers who send and recieve them?  In other words, will

the market be supply driven or demand driven?  The telecom revolution is

being fought over this very question.

Certainly the telcos WANT the network to know so they can charge according

to what people value.  The rub is that the costs of delivering bits are


Let me put it another way.  TV over Internet is around, say, 6 Mbits. Or

about 1000 times voice.  If the network doesn't know the difference, then

the two extreme pricing scenarios are (a) to charge as if everything is

voice, about $5.00 a minute for TV, and (b) to charge as if everything is

video, which would make voice free by any current video pricing model. Of

course there are other pricing models, but these define two important

corners of the content-agnostic network.

AT&T is touting "facilities based competition." In other words, it wants to

control the network from the core to the set top box. One of the main

reasons for this is so it can know the content type at its origin -- if it

comes from the video hole, you can charge differently than if it comes from

the telephone hole.  And you can charge a third price if it comes from the

text-terminal hole.

Furthermore, one reason why so-called QoS, quality of service, is such a

critical telco issue is because most QoS schemes require some knowledge of

content so the network can treat e.g., voice different than e.g., data.

Once you do that, the "network knows" and the telephone company charges

differently for different content types.

Negroponte is correct that the current Internet can't tell voice from email,

but there is no reason to assume that it will always be that way. There are

some powerful forces working to turn the situation around.  I hope that the

content-agnostic network triumphs -- it is one of the characteristics that

have made the Internet such a great success so far.



October 10-17, 1999, Geneva, Switzerland. TELECOM99, the every-four-year ITU

extravaganza that always seems to surprise AT&T's leaders.  I will be there

10/9 through 10/13 posing as a wild-mannered columnist for America's


October 27-29, 1999, New York City.  The New Economy Conference, with John

Browning and Spencer Reiss.  Plus an array of people who believe that

knowledge is wealth, that bigger ain't necessarily better, and that there

are more opportunities and fewer guarantees at the edge.  I don't know what

I'll be doing there yet, but I suspect I'll be there.  Watch for the emergent agenda.

November 4, 1999, New York City.  "TechBrains Seminar" with Merrill Lynch

Technology Advisory Board members.  Featuring Phil Neches (founder of

database machine company Teradata), Don Norman (who wrote "Turn Signals are

the Facial Expressions of Automobiles," and other worthwhile reads), open

source spokesman Eric Raymond (who wrote the must-read essay, "The Cathedral

and the Bazaar"), and several others, no less distinguished, whose work I

don't know as well. I'll participate too.  Email me if you are seriously

interested in attending.


COPYRIGHT NOTICE: Redistribution of this document, or any part of it, is

permitted for non-commercial purposes, provided that the two lines below are

reproduced with it: Copyright 1999 by David S. Isenberg -- -- 1-888-isen-com


[to subscribe to the SMART list, please send a brief,  PERSONAL statement to (put "SMART" in the  Subject field) saying who you are, what

you do, maybe who you work for, maybe how you see your work connecting to

mine, and why you are interested in joining the  SMART List.]

[to unsubscribe to the SMART List, send a brief  unsubscribe message to]

[for past SMART Letters, see]

[Policy on quotes: Write to me.  I won't quote you without your explicitly

stated permission. And if you ask for anonymity, you'll get it. ]


David S. Isenberg, inc.  

1-888-isen-com            1-908-654-0772

** -- the brains behind the Stupid Network


Date last modified: 14 October 99