Tim Oren's Due Diligence

John Patrick on WiFi vs.

John Patrick on WiFi vs. WiMax

Another excellent WiFi post from John Patrick, apparently brought on in part by some press confusion between that and WiMax (802.16). He gets it right: WiMax doesn't threaten or replace WiFi, they solve different problems. However, variants of WiFi such as beamsteering or meshing could be repurposed for premises links and displace part of the need for WiMax, if the latter market takes off slowly.

After thoughts on Isenberg's Stupid

After thoughts on Isenberg's Stupid Networks

[NB: I'm back after a (believe me) well deserved three day weekend vacation, and a day lost to some strange behavior by OS X that also munched up one post before it upstreamed from Radio. System now apparently stable again, so here's a redo on that last conference story...]

I've known David Isenberg since before he bolted from AT&T Labs, and agree with most of his Stupid Networks thesis. We also share views on some of the amazing miscalculations recently emerging from the FCC. Having said as much, though, I'm going to take him to task for a few points in his ETech talk that could be better thought through for business outcomes.

Part of the pitch is now built around logic by Roxane Googin who apparently had some prescience regarding the telcom and market implosion. It is certainly the case as both propose that you can't sustain a carrier business model that includes capitalizing switching equipment on 20 to 30 year schedules backed by debt. At least not in any world that persists in innovating around you. All you can do is prop up a polite financial fiction that diverges further from reality each day.

It's going it too far, however, to say that "any system that is totally generic has no barrier to competitive entry. You can't differentiate yourself." Someone had better call up UOL and Earthlink with the bad news so they can get on with folding their tents. Obviously there are other ways to play the game. Even a rather basic treatise on positioning will suggest that perhaps customer service (what a thought!) and image branding might be options. Taking a look at those two ISPs might suggest such a thing could be happening. If we can get along with the inevitable writedowns, rather than defending the indefensible through regulatory backfires, other telcom sectors might manage as well.

How to deal with the current underavailability of backbone to home&office LAN connections? David's solutions chart includes a combination of regulated monopoly (meet the new boss, same as the old boss), government ownership, and customer ownership. There's undoubtedly some missing, given the 'no business here' fallacy, but I like at least one of these anyway.

It's not government ownership. It's great that at least Stockholm seems to have got it mostly right, but I can also point you to some dirigiste fiascos that set their own causes back by years. The political risks inherent in making such a long term and overt commitment of public resources ensure it's not going to be a fast moving process at any rate. A poor place to look for market drivers.

Customer ownership has a lot to say for it, however. As Bill Gurley pointed out in one of his ATC letters, WiFi is letting customers effectively substitute their own capital expense for that of 3G carriers. We're on the edge of getting both 802.16/a and variant WiFi implementations that can run a decent medium weight backbone to premises link. And they will be appearing at prices in the $300-500 range per connection. That's not far above the $150-250 pricepoint that even consumers proved to be willing to repeatedly pay for improving bandwidth during the escalation from 1200 up to 57.6k modems.

Yep, backhauling that'll be one of those impossible 'undifferentiated' businesses that supposedly won't work. But would you rather be an incumbent with all those switches and all that copper on the balance sheet, or be starting with your customers carrying a lot of the financial load and technology upgrade risk? As I've mentioned before the regulatory battleground that matters right now is making sure there's sufficient spectrum available that this path can't be blocked, and we can get the scale economics working. The rest is a rear-guard action.

Nano-nano: Jurvetson and Drexler I've

Nano-nano: Jurvetson and Drexler

I've held off blogging Steve Jurvetson's session from yesterday until I had Drexler as a sort of counterpoint. In case you don't follow the VC trade, Jurvetson is sort of the marquee nanotech proponent in the venture capital world. Unsurprisingly to the bloody-minded set, he's looking first for near term business implications in nano projects. It's best if they combine here-and-now pragmatism with an ultimately disruptive outcome when and if they succeed.

Steve lays out two paths ahead. The 'top down' path fits nanometer molecular features within an existing manufacturing framework, specifically that of planar silicon processing. In places where it is hitting its limits, some very specific techniques for self-assembling (for instance) molecular memory or interconnects could break down road blocks, without having to build up an entire novel infrastructure. Two observations:

• He talked about the issue of scale, meaning the problems of connecting into molecular scale structure inserted within a macroscopic world of flipchips and pinouts. I was also thinking about 'scale', that the top down path is a bet on leveraging the existing scale economies of silicon processing.
  
• Jurvetson showed - and Drexler cited - a Moore's Law graph looking back to mechnical pre-computers and extending to now, of course with the implication that nanotechnology is how the curve will be continued. Interestingly, Steve then dropped the topic - none of his examples, and none of his ventures, are directly about squeezing out more cycles. It's getting increasingly hard to get paid for riding that side of the silicon learning curve, it may also be a weak reed to support nanotech. Watch the man's feet and wallet, which seem to be betting on energy, sensors, displays and memory, not cycles.

Back to Jurvetson's second path. Unsurprisingly, it's 'bottom up', in this case up from biological goo. We're talking about hijacking existing nanotech libraries cooked into the genetic codes of organisms, from twiddling heat shock proteins in order to create a regular self-assembly grid for circuits, to forced mutations of bacteriophages toward externally imposed evolutionary goals. Scale in this case comes from exploiting a massively parallel genetic algorithm - the original!

While he didn't quite say it, I'll infer that Steve sees these two paths converging in the middle. Leverage the scale market from the top, leverage the evolutionary approach from the bottom. Borrow from both Gordon Moore and Mother Nature, and keep your capital light. Anyone who's played with GAs knows that specifying the problem correctly (just like neural nets and linear programming) is often the better part of the solution. Are we heading toward a future in which the specification of an interesting 'target' nanoscale engineeering program becomes tantamount to its solution? He seems to hint at it with the notion that some types of nanoventures may be analogous to drug discovery, both in intellectual property and business structure.

If Jurvetson is capital light, Drexler is still Big Science, or wants to be. I first heard a version of his pitch under the trees at Camp Swig, probably at Hackers 2.0, which would make it 17 years ago. It hasn't changed in essence, but it's grown darker. We're not only to worry about building the theory and infrastructure for self-replicating nanomachines, but buy a regulatory insurance policy against Grey Goo cooked in the outback by terrorists. Get your arms race here! I guess I recognize the business model - it's called 'start a war and sell arms to both sides.' Except, the guvmint itself seems to be stubbornly funding pretty much here-and-now applications. The guys on our side of the hypothetical arms race are building point applications that keep us ahead of the Bad Guys, and dissing the Big Nano approach.

You've got to watch out for Kevin Kelly, who asked the most subtle question of the Drexler session: "What have you learned in 20 years?" The answer is apparently that bureaucratic scientists persist in not seeing it Eric's way. The Coyote persists in using a different set of physics, and stubbornly just keeps walking on air. I'm sorry, this is not a useful learning curve. Devastating. (Note to self : refrain from calling on KK if I'm pitching...)

The Old is New Again

The Old is New Again

Kottke succinctly captures a mood of deja vu that I've also felt this week at E-Tech.

Dan Gillmor on audio blogging

Dan Gillmor on audio blogging

I can't recall if this came from the journalism session or the warblogging panel, but this POV struck me:

"Audioblogging... I don't get it. Posting multimedia is fine, I love it, but...
Exposition in audio is hard, most people don't do it well, it takes longer...
I'm going to audio post the worst press release I get - real aloud with emotion..."

Kind of echoes the way I feel about both audio and v-blogging. Great for experiential type blogs, but not good for exposition. Both are wretched summary media. I mostly dish out technology and industry rantings analysis here, text is the right vehicle.

Blog wallpaper This is cool....

Blog wallpaper

This is cool....

Via Jeff Jarvis

Clay Shirky at ETech A

Clay Shirky at ETech

A few tidbits that resonated with personal experience or are relevant to topics I've been hashing over here (you may infer that I agree with them):

• General purpose reputation systems are useless. They are not portable between contexts. People who would cheat on their spouses might not cheat at cards (and vice versa). Ebay did us a disfavor by creating a 'reputation' system that actually referred to atomic economic transactions - it doesn't really translate. Real reputation is emotional. The penalty for identity surfing is and should be leaving your reputation (friends) behind.
  
• In the group software environment, some things must be hard/limited, else the core group cannot defend itself.
  
• The downside of Metcalfe's Law: Since the potential connections grow as the square of the group size, the density of conversation falls - and eventually sputters out. There must be a way for the group to defend itself against scale. Note relevance to the discussion about ecology of networks.
  
• The impact of immersive environments was overestimated, not as universal a pattern as first thought. Many early studies focused on what was different - novelty effects in MUDs and other realtime contexts - LambdaMOO and Habitat, for example. Non-immersive software has more long term impact. The patterns is small pieces loosely connected.

(It's also great to hear from someone who remembers CommuniTree!)

Alan Kay's talk: Smalltalk still

Alan Kay's talk: Smalltalk still stands alone

The first half was a reprise of early history - Licklider, Sutherland, up through the Alto and 'Dynabook'. Seen this before, many times. Have a videodisc with a lot of the material burned on it somewhere. Worth doing, since every new generation needs exposure to this stuff in the original form; it always loses something in the retranslation of each computing generation. And if people won't do their homework, this is a way to stick the prior art right in front of their noses.

Now onto the latest Smalltalk incarnation. Cool of course, great that we can finally get that much horsepower into a laptop. Shades of things I worked on at Kaleida Labs and hoped (vainly) for at Electric Communities, nice to see it come to life.

Now the skeptical take. It won't make any difference in the current generation. The insistence of a pure system from the bottom up allows clarity and cleanliness. It also means that Squeak, like the original Smalltalk and most incarnations in between, "doesn't play well with others". The real world of computing is a complex ecology of platforms and vested business and architectural interests. Doing a top-to-bottom standalone - and making a market success of it - is one of the hardest acts out there, particularly when it appeals to an interest (programming) that is a relatively minor taste of most users outside of this conference. Here's a fearless prediction - maybe I'll survive long enough to be held accountable - it will take two more major generations of computing equipment before this vision happens, and when it does, it will appear in childrens' toys, which can be standalones.

Esther joins the party! Come

Esther joins the party!

Come on in, the water's fine! You don't have to rat on your portfolio companies, but we do expect an annotated guide to every swimmin' hole on the continent.

802.11 based meshing/routing At the

802.11 based meshing/routing

At the FHP Wireless talk. I know this domain pretty well so I'm not going to plow over old ground. If you think it's interesting, in addition to FHP, check out:

• Mesh Networks
  
• KarlNet
  
• Tim Pozar's homebrew efforts at Bay Area Wireless Research Net

Here's a slightly aged Kevin Werback article about the topic. Robert Berger's Ultradevices company referred there is dead, but he's at the conference so track him down. Also check out the meshing capabilities in IEEE 802.16a.

Now it's getting a little more interesting, since audience members who are working on the same problem are banging away with questions, attempting to extract some of the protocol secrets that might be in the FHP architecture. Getting more context from the audience than the talk...