e$: Walking "Down" the Hierarchy

Robert Hettinga
44 Farquhar Street
Boston, MA 02131

Comments: rah@shipwright.com


Boston, Massachusetts


Practically my only rant these days is about how the exponential collapse of microprocessor prices has mutated our communication structures, and thus our social structures. This creates interpersonal and organizational networks which are, as George Gilder likes to say, "heterarchical", instead of hierarchical. The interesting thing is, if you look back in history, we didn't really have the ferociously large multilevel hierarchical organizations we've all come to know and love until communications technology started to accellerate, beginning with railroads, through telegraphy and telephony and ending with mass media like radio and television.

The Catholic church in the Middle Ages, for instance, was only three levels deep. The Roman Empire, slaves and all, was never more than 4 or so.

China, the possible exception to this, had quite fast communication. Post riders in long relays kept the country knit together very tightly. Genghis and Kublai Kahn's mongol horsemen, for instance, could get a message across most of Eurasia in less than a month. Cities, creatures of the invention of writing, could be more heirarchical than villages because comunications channels were short, lots of people didn't know each other intimately like they would in a village, and there had to be some way to "switch" all this information. Hierarchies are a good way to do this. The course of civilization (literally, the use of cities) has been, until now, a continuous trend towards larger and more integrated hierarchy.

The industrial revolution threw coal, and then gasoline, ;-), on the fire of hierarchy. Our big modern hierarchies, like the Fortune 50 and the U.S. federal government, were built on increasingly faster communication, though all this information was still being switched by human beings. Nowadays, things are beginning to go the opposite way, walking back "down" the ladder of heierarchical evolution, because switches, especially switches of information, don't have to be people anymore. They are, for the most part, microprocessors. The first person to actually talk about this collapse of hierarchy as it pertains to the telephone network was Peter Huber, who called them "geodesic" networks, in honor of Buckminster Fuller's discovery of geodesic structures in nature, and Fuller's architectural designs based on those structures. In other words, the resulting public networks of today look just like Bucky's domes, if you were to "flatten" them into three dimensions. Huber's famous line for all this is "Nodes are now cheaper than lines". This cheapness of nodes creates an environment where smaller and more powerful microprocessors break information into smaller and more useful pieces. My favorite analogy has been to think of geodesic networks as information "surfactants", fractally breaking up blobs of information on the internet much the same way dish soap fractally breaks up grease blobs floating in the sink.

Oddly enough, all of this comes to mind because my ISP, The Internet Access Company (TIAC), of Boston, is having growing pains again, and I had lots of time to kill recently while they fixed the new RAID on their mail server. In my stranger moments, it seems to me this company, which has done so much to create this process of dissolving hierarchy, is in the process of being surfacted itself. The paradox is, by surfacting into more functional units, TIAC could make much more money for its shareholders someday, the same way that the breakup of AT&T inadvertantly enhanced the value of it's shares more than a decade ago. To do that, we have to talk about lots of things, not the least of which is how we're going to work in the geodesic markets created by these networks.

People are starting to create geodesic businesses these days. Their "jobs" are being broke up into a series of independent projects, each creating an ad-hoc team of people to solve a problem. When the problem is solved, those people go their separate ways. Originally operating this way in corporate and academic structures, the internet has allowed people who work with their heads this way to effectively freelance in extra-corporate teams, to collaborate on something, then to collect their money, and to market themselves for another project. I work this way. Lots of people I know on the net work this way, and I expect that more people will work this way as time goes on. I'm almost convinced that the only reason we have "companies" as we now have them in the software business is because we need mechanisms to interface with the more industrio-hierarchical institutions in the rest of society. A form of financial and legal protocol conversion.

At the turn of the last century, the two largest job categories were farmer and domestic worker. By the middle thirties, the largest job category was factory worker. These days only 2% of people farm, and less than 20% actually make things. The rest of us are "service" workers. "Service" is a category meaning "not manufacturing, and not farming". I believe that someday, we'll have .2% farmers and 2% manufacturing. Lots of the "service" jobs usually involve specialty retailing or "retail manufacturing" of some kind, in the sense that MacDonald's "manufactures" food, or a computer sign-maker franchisee manufactures signs. I think that with enough processors and bandwidth, the net surfacts many of those physical storefronts away, removing a significant amount of cost of doing business in the process. The signmaker doesn't need to buy expensive space in a mini-mall or on the interstate's business loop. You order the sign from her directly on the net, and she delivers, or more likely has it delivered, to you, though probably not by the aggro skatepunk Neal Stephenson showed us in _Snow_Crash_. While we're on the topic of delivery, Neal Stephenson, and his pizza-delivery king, Uncle Enzo, it could be that people working on the net at home might prefer delivered food, if it was as cheap as any other fast food. Also, as the ubiquity of processors make kitchens much smarter, they might just want ingredients delivered. Remember what's happened with microprocessor controlled home bread making machines lately. Maybe Mrs. Jetson's kitchen wasn't that far fetched after all.

Shaw's supermarkets in Boston has reinstituted grocery delivery, for instance, with the World Wide Web as a front end. Shaw's service is still quite lumpy right now. The list of things you can buy in a grocery store is huge, and Shaw's insists on sending you their entire product list (measured in megabytes), so that you can choose what you want in a quasi offline fashion. It almost seems to me like grocery inventory "shovelware", where they've dumped their inventory architecture straight to microcomputers because the modern microcomputer is almost as powerful as the mainframes this code was written on originally, say 30 years ago. Nonetheless, this is nothing that bandwidth can't fix in the the long run, and I think ideas like this one are eventual Wall-Mart killers. Not to mention killers of large grocery chains like Shaw's.

Wall-Mart and Shaw's are essentially a forward storage mechanism -- warehouses, in other words. While economies of scale are achieved this way, the net allows you to segment products by content, and not geographically, which is what Wall-Mart's forward storage model is built around. When you can search for products with agents which know what you want, buy them directly from the manufacturer, and have them shipped overnight or even delivered the same day, the need to store a diverse inventory, well, surfacts away. Things become more price competitive, and, at the same time, more customised, which, while seeming like a paradox, are both a result of collapsing processor and bandwidth prices.

It's an old story in markets everywhere, a market becomes price competitive, forcing people to differentiate or die. On the net, this can happen much faster, because you are directly linked to your buyers. Geography and time are not nearly the factors they are in phyiscal retailing. As soon you write something like an article or software, you put it on your web-server and it can be sold. The almost weekly addition of features to Netscape beta versions last year is a good example. You could almost see "traditional" geographic-distribution software developers, like Microsoft, shaking their heads in dismay. What all this breeds is differentiation by function.

This will not be limited to software and wetware. It's easy to see where this could happen with hardware as well. If you design something new, you can ship a prototype, or more to the point, email the design files, to a manufacturer with a drop-ship arrangement, put a discription of it on your web-page, and post an announcement where people are looking for it. It's possible to imagine a "syndicate" of manufacturers for any given product, people who have manufacturing facilities closest to the particular consumer making the order. Certainly it's possible to imagine numerical machine controls evolving to the point where manufacture is a mostly commoditized function, in the same way that agricultural produce is. That requires, of course, an economic demand, and frankly, Alvin Toffler's "mass-customization", the idea of the general-purpose robotics shop doing manufacturing piecework next door ala Japan Robot Lease Inc., failed, because information was still centralized in manufacturing firms. All the demand for production capacity came from one place, the company marketing the final product.

The direct, geodesic connection between buyers and sellers on the net surfacts that information away from a manufacturing hierarchy, and thus resuscitates Toffler's mass-customization idea. With this kind of manufacturing syndicate model, an auction market for the production of various manufactured products is created in your geographic area. Raw materials are bought geodesically and shipped as close as possible to the consumer. Materials and people converge on local equipment to manufacture products as needed. When a product is hot, batch runs are possible, but it's possible to see production happening in smaller and smaller lots, eventually one at a time. When looked at through the lens of the geodesic network, it's possible to see that the whole idea of just-in-time manufacturing was the beginning of what may be called, well, geodesic manufacturing.

As manufacturing gets geodesic, "jobs" become scarcer, and one-shot "projects" increase, just like in the wetware and software businesses. However, overall, fewer people make things for a living over time. The paradox about this is that those people probably won't be fired, so much as attracted away, by software and wetware businesses. This is pretty much the same way that manufacturing attracted people off an increasingly mechanized farm with higher paying manufacturing jobs in urban areas. Since quality of life becomes more of an issue, people start to move farther and farther from urban areas, the means of production will probably follow them, and transportation will follow manufacturing.

For instance, cars are a rudimentary geodesic transport mechanism. Suburbs and edge cities are a rudimentary creation of that form of transport. With the net allowing more people to work where they want to, just because they want to live there, demand is created for more and more geodesic forms of transportation, transportation where the "switches" aren't people: autonomous vehicles. Already there are several autonomous-vehicle microrail proposals in the works, and one is under development in a Chicago suburb, a pilot project of the RTA and Raytheon. If location-transponding chips were cheap enough to imbed in roadways themselves someday, putting electricity there as a source of power for locomotion wouldn't be far behind, and the result would be roads full of autonomous vehicles moving people where they want to go, and goods where they're needed. Multimode hybrids with more traditional forms of road, rail, and air transportation are certainly probable, much the same way that container ships and piggy-back railroad cars interoperate with trucks today.

Give up? What does this have to do with my beleagured internet service provider, TIAC? Well, in a word, reliability, which is why I'm sitting here waiting for their mail server to come back up. In two years, the ISP market has gone from a market where private ISPs had to offer everything, because nothing was there, to where the traffic on all of their services is so high that they can't do it all anymore. In the same time, demand for reliability has skyrocketed, because a whole lot of people are starting to use the internet for geodesic businesses, like mine. I get about 300-500 mail messages a day, and I distill them down to about 30 or 40 or so, to be sent back out to my e$pam subscribers. In addition, I have consulting projects where I may be working in a team with 4 other people, like the one I'm working on, and I generate the occasional article, or rant, like this one. Fortunately, if TIAC's mailserver is up once a day, I'm golden. However, when things are hopping, like when the President of Netscape forgot to wear his asbestos Y-fronts and said he thought GAK (Government Access to Keys) wasn't that bad an idea, events require a little more than one perusal daily. I expect that it's going to get more like that in the future.

This happened with cars, by the way. Cars started out as toys that mechanics built for fun, then toys that mechanics built for rich people, then toys for the rest of us, then useful tools that everyone should know how to use, and finally cars became so necessary to everyday life for so many people that they don't require tune-ups but every 100,000 miles or so, and it takes a specialized technician just to change the oil.

Notice that we're just passing the "toys for the rest of us" phase of the internet, and rapidly moving into the "useful tools" phase. Not wanting to beat the analogy like a dead muscle-car, but did you notice that cars became utter necessities just after interstate highways were built? In the same vein, it's only really been 10 years or so since we realized how really useful a thing the internet is, much like they way we discovered the usefulness of paved roads in the 1920's. Things are happening much faster on the internet, but we're still just barely in the "Route 66" phase of the internet. So, it's still pretty early to tell, but, contrary to the current popular consensus, I don't believe that passive, extremely asymetric applications of bandwidth, like video-on-demand, will be the "interstate" builder everyone thinks it's going to be. I think it will be something more important than that. Money.

As Dan Eldridge of Digicash says, "It all boils to money on the wire." When I can hire you, and pay you for the work you do for me, even in cash, no matter where you are, and you never have to go anywhere to do the work, including never having to leave your home to collect your pay, or even to spend it with anyone else, that's a strong decentralizing economic influence. People can live anywhere they want to. Resources they need to live will follow them wherever they go. If you want to live on the North Fork of the Flathead River in Montana, or on St. John in the Virgin Islands, or on a circumnavigating trimaran, or even in the small african village you were born in, the work will follow you there. So will the necessary telecommunications, water, waste disposal, transport, and everything else you need to live there.

All of this implies an absolute reliability in the network, and, as network people know, reliability is redundancy. Two packets in the same message don't necessarily take the same path. The route depends on the network's load. Lots of little vendors of the same kind of service, sharing resources in demand-based syndicates, may be better than one giant server-in-the-sky. A single node can't switch all the available bandwidth. It chokes.

At first blush, this doesn't bode well for growing centrally-organized service providers of internet services, especially small ISPs like TIAC. I say small in TIAC's case, because even though they've got just shy of 15,000 subscribers, not bad for two years work, they're gnats to the likes of AT&T, which just entered the ISP market this week.

Still, I think that TIAC's in much better shape than AT&T. One of the strange things I heard the C|Net guy say on Leherer's PBS NewsHour was something to the effect that since AT&T was going to be where you get your mail, say, rah@att.net, and that would lock you into buying your long distance service, much less your internet access, from AT&T. I don't think so. Remember our amazing surfacting geodesic network, folks.

First of all, my inbound and outbound mail servers can be anywhere. They can even be at TIAC, for that matter. Same with my web-server, which in my case is a borrowed machine on borrowed bandwidth on the other side of the country from me. News, domain names, everything you need to work on the net can be gotten from the most efficient, reliable competitor in what amounts to a global auction market. At first blush this looks like a large but specialized commodity market. Of course, we know that it's not a commodity market. This is the segment-or-die internet, right? We just don't know what the next level of functional segmentation is going to be, even if the current level is plain as the nose on our faces.

So, looking in the mirror, as it were, the first thing TIAC should do, with a monster like AT&T jumping into their business, even if their business is growing so fast that nobody can keep up with demand, may be not be for TIAC to grow at all, but to break up into smaller pieces. Or maybe at least break up their product line. Currently, TIAC differentiates their service, for the most part, by the number of bits you hold an option to, in some fixed combination of the bandwidth of your connection, or the time you spend online. It seems to me that if they offered an ala carte menu of mail, news, web and domain name service, that they could then turn around and market that to the whole net, not just the customers showing up on their portmasters. They could more easily price these services according to the demand, and thus meet the demand for those services reliably. It may even be that someday, TIAC wouldn't be in the connectivity business at all, and leave that for the AT&Ts, Bell Atlantic/TCIs, or USWest/Continental CableVisions of the world. Heck, maybe internet dialtone will be so cheap that small local utilities would handle it, and the big players won't be able to compete. Besides, digital dialtone may not be a business TIAC wants to be in, when there are better value points in more differentiated internet services. Just like I laugh when I see Citicorp saying it wants to control the digital bearer certificate business, I laugh when I see AT&T getting into the web- or mailserver business.

Once you've differentiated your business on actual markets for services, rather than bits, it's easier to spin them off if they get profitable, or cut them off if they aren't.

For some reason, all of this reminds me of a Francis Bacon quote on innovation: "He that will not apply new remedies must expect new evils; for time is the greatest innovator."

In other words: In a geodesic market, you have to differentiate, very quickly, or die.



Bob Hettinga


Updated: March 2, 1996


[ Back to the previous Rant ] | [  home page ] | [ On to the next Rant ]