Lightweight speed and power at Emerging Tech conference

by Andy Oram

Related link:

This morning at the
Emerging Technology Conference,
O'Reilly staffer and author Rob Flickenger explained why
we need wireless community networks (the subject of
his recent
and how to make one that is flexible enough to provide great bandwidth
to its owners while still offering public Internet access to anyone
who happens by. I also saw a way to build Web applications that is
just as lightweight and unimposing as Rob's wireless network.
And I threw in a visit to
for good measure.

An illustration of the strength of the Emerging Technology Conference

Combining one's resources with neighbors to provide a local network is
much cheaper and more effective than making each individual buy his or
her own Airport and set up a personal network, particularly when
people share an apartment building and don't want to trash each
other's broadcasts. Rob's
exemplifies what one can do with cheap, widely accessible equipment.
Response to his talk was impressive; nearly fifty people filled the

Rob laid out an expanding range of configurations:

  • Simple "hot spots" (providing public

  • Point-to-point connections (which let people reach friends on
    the other side of town by hopping from one access point to another)

  • ParaNets, which involve a mesh of wireless access points and
    effectively provide a network parallel to the telephone wires.

The latter two configurations are particularly interesting (here I am
switching to my personal observations) because they represent a new
movement called "asset-based facilities." It's a reaction on the part
of the users to the sluggish and sometimes abusive expansion of the
traditional, large telecom players. Users empower themselves by buying
their own assets instead of leasing them; they can also create their
own point-to-point and mesh networks that permit lots of productive
data exchanges without mucking around on the Internet at all. Rob
pointed out that a key strength of ParaNets is that each access point
is owned by someone who has a vested interest in making it work well.

This movement also illustrates what I pointed out in
yesterday's weblog
when I wrote that by one could understand each trend promoted by the
conference better by learning about and combining all of
them. Wireless or wired facilities that connect to each other in
hop-to-hop fashion are hardware versions of peer-to-peer networks such
as Gnutella. As such, their architects have a few things to learn from
the peer-to-peer movement.

For instance, users that depend on intermediate sites to carry their
traffic should find ways to recognize rogue sites that drop or corrupt
the traffic. Related reputation system concepts will be presented at
the conference tomorrow by an author who worked on the book


Supported by several colleagues, Rob also presented an intriguing
wireless firewall system called NoCatAuth. It cleverly allows
different classes of users (the owners, collaborators of the owners,
and the public at large) get access to the network with predetermined
rights and guarantees of service.

NoCatAuth is basically an SSL server; one can either become one's own
Certificate Authority and deal out access to known, preregistered
users, or offer service to the general public by getting a standard
certificate from a well-known CA. In either case, users can connect
with the assurance that names and addresses are not being spoofed;
users in turn can be kicked off the network if they abuse it with spam
or other unwanted activities.

To sum up, I find it amazing that wireless networks have succeeded at
all--not to mention having become the new, exciting grass-root
Internet movement. Wireless networks use junk spectrum (because it
does not require an FCC license), can be easily blocked by trees and
other barriers, suffer from security standards that are seriously
flawed (although Rob says they'll be improved), and get bad press like
"Wireless users steal DSL access."

But succeed they have. And this is because they are incredibly cheap,
incredibly fast, and almost totally unregulated. Some experts think
users can cooperatively work out continued growth without regulation,
but I think there's a lot the FCC could do. Rob mentioned his hopes
that ultra-wideband (UWB) would be approved by the FCC in a year or
two. In fact, it occurred to me that the success of wireless may drive
renewed adoption of DSL and other wired technologies--anything that is
free from the crippling service-level agreements of cable modems.

Navigating Yahoo! nonvirtually

I had to miss my colleague Brian Jepson's talk on .NET Web Services,
where he apparently spun an elegant and powerful programming interface
to a commodity trading system, because I was scheduled to have lunch
at Yahoo! with author Jeffrey Friedl of Mastering Regular Expressions

Yahoo! has put up a large and attractive modern facility right on the
bay. Their contribution to the environmental movement apparently
includes designing a parking lot that fits only compact cars. The
engineer who started the company still occupies a cubicle like any
other employee, although his is a bit larger than most and faces a
view of the mountains.

I was impressed with the company spirit that apparently holds strong
at one of the companies to maintain its vision and its cash flow (I
consider O'Reilly & Associates another one) through several
generations of fast-moving Internet time. To a large extent,
incidentally, Yahoo! runs on O'Reilly books. You can find quite a
library in nearly any cubicle.

Making Web Services natural

Leaving the site that best exemplifies the 1990s Web to come view the
emerging twenty-first century version, I got back to the conference in
time to see James Duncan Davidson (yet another O'Reilly author) build
an application on top of Brian Jepson's raw classes. Duncan frankly
admitted a long-time suspicion of attempts to pump up the Web into a
complex application platform, but said "As time goes on, it's apparent
that something really is there with Web Services."

Still, for this presentation, he eschewed "huge frameworks or huge
toolboxes" like SOAP and WDSL. Instead, he built a Web service using
human-scale building blocks: XML to hold the data, JDOM to extract
this data easily into Java classes, Mac OS X's Cocoa to build a
beautiful user interface. And (let us not forget) a simple HTTP GET
for the "heavy work." Those who disapprove of using GET for this kind
of operation can use POST instead.

JDOM is a recognizably DOM-like interface that deals with the usual
hierarchy of XML tags and contents. Cocoa (based on NextSTEP) has
apparently created a programming system that lets average programmers
use the Model-View-Controller method robustly without having to be
software engineering whizzes.

During the discussion at the end of Duncan's talk, one of his colleagues defended SOAP. But all participants agreed that proper Web Services were lightweight and simple. Grand schemes like transactions have no place in them.

What technologies will really make it big?