[ Back to the previous Rant ] | [home page ] | [ On to the next Rant ]
Robert Hettinga
e$
44 Farquhar Street, Boston, MA 02131
Comments: rah@shipwright.com
Boston, MA
12/29/95
First a little administrivia. Going forward, I'm going to be posting my rants here first at e$@thumper.vmeng.com, with a forward from e$ of the given rant, after some arbitrary delay, to any other lists that might be appropriate to the subject. My sponsors have paid for this list, and the readership of the e$pam and e$ are there to hear what I and others have to say about e$ and its consequences here, and so my stuff goes onto the e$ list first from now on. Like all my writing on the net, redistribution with attribution to me and the other people I cite (When I remember to cite them all...) here is just fine. If you're publishing it, and you pay your authors, I'd like to be compensated too, of course.
Not that what I say here is going to be all that earthshaking, but I thought I'd clear the air a bit.
I've read a lot of interesting things about the future of the net in the last few weeks, some of which I've sent on to e$pam, and I'd like to talk about them, by way of clearing the decks for the New Year.
First, I'd like to go revise my model of net.reality a bit.
Most people look at the net as a hierarchy. Architecturally, it is exactly that, from the IP addressing scheme, to the object super-hierarchies in component software models, to server-mirroring, to just about any kind of structural component of the net you would care to look at.
Physically, big lines get broken up by big switches into smaller lines which get broken up by smaller switches into smaller lines, in a fractal process which ends up at you or me, where it goes back up each larger level to get where it needs to go.
In software, the CORBA object model, and the super-object-model that the research people have been cooking up at Microsoft, all have a "root" somewhere; the "top" of their taxonomical system, if you will.
In information, there's a source of the information, and it gets accreted with other information and synthesized and averaged and summarized and rolled up into some larger aggregate which allows you to have some knowlege about that information and other information like it.
In finance, my money gets pooled with other people's money through several larger aggregation layers and then invested or spent centrally somewhere. The recipient of capital or cashflow then spends or invests it in fractally smaller chunks until I get it in a check for something I do.
I could go on and on with this, but you can see my point, and it is the same point Rich Lethin loves to use on *me* when I start talking about "geodesic" anything, that is, the idea that Moore's law exponentially collapses switching costs, making nodes cheaper than lines, making the network and the software processes mapped onto it more geodesic instead of hierarchical, and "surfacting" information and software into fractally smaller and smaller pieces.
What I've been saying to Rich, particularly when I talk about geodesic networks, is that the message itself is point-to-point, even though the actual electrons may flow more or less hierarchically around the network. That's kept him busy while I made my getaway. It always felt like a sophistic shuck, myself, but I'd learned to live with it until now.
I don't know if I've gotten anywhere, but I've been thinking about it a bit, thrashing anologies from other parts of the world, --the major way I think, unfortunately -- to describe what I see out there. I've been thinking about biological models, because in my stranger moments, I like pretend that the net is an electro-biological entity.
For instance, the circulation of blood is a good anology, I think, because all the endocrine messages in the blood stream are ultimately broadcast from a single cell and paradoxically sent point to point to another cell -- just like things are on an ethernet wire, or on Gilder's fanciful dark fiber, even though the circulatory "backbone" is hierarchical.
The most obvious example of course is the organization of neurons, in that the brain pathways are essentially geodesic, but we still have to deal with the hierarchy of nerves outside of the brain.
The one thing I think that differentiates these models from the hierarchies we encounter in social life, the industrial induced control hierarchies we all rail against, and the stuff I like to quixotically do battle with on the net, is that every one of those biological hierarchies is chaotic. There is no pretense of top-down control of the system. The load of the system is hierarchical, and so the system organizes itself hierarchically. There are physical forces which create physically hierarchical stuctures, but they're usually set up to solve the problem of efficient distribution of something over a distance, like draining a watershed, or getting blood to a central heart and lungs, or nerve impulses back and forth to the brain.
When distance isn't a problem, networks, like the brain, tend to get more geodesic. Bandwidth is maintained by an abundance of neuronic "switches", doubling as processors, tripling as memory, each with some number of connections to other neurons, rather than a bunch of "fat" nurons doing all the signal processing. As an exception which proves the rule, note there *are* in fact "fat" neurons, more precisely redundant neuron pathways, particularly between the two halves of the brain, and between the brain and the rest of the body.
So, what else is new? We still have hierarchies on the net, right? We're about to bump up by many orders of magnitude the number of possible IP addresses real soon, so that someday your toaster can tell your alarm clock to wake you for breakfast. I've ranted, tounge-in-cheek, about the "dangers" of the "X.blabla" book-entry view of the world, with hierarchical, government-as-root certification "authorities", and the consequences of having an audit trail on your every net-based financial activity. Most of this X.blabla stuff will come to pass, mostly because it's the easiest thing for the financial system as it's currently organized to do. It's sort of like financial "shovelware", moving the contents of one financial medium, the hierarchic industrial paradigm of government regulated central banking systems onto the new medium of the internet.
However, in a world of micro-pay-as-you-go packet routing, where routers may someday spot-auction their bandwidth on a demand basis at packet prices displayed best in scientific notation, all those audit threads could lead to a Gulliverian restraint on personal freedom, much less on individual privacy.
Fortunately, I don't think that's going to happen, because those same Lilliputian audit trails will just get in the way and slow the system to a standstill. We need to get more chaotic.
I think my contention on this is that as we get smaller and smaller, the more chaotic it's going to have to be. Book-entry based transaction processing systems will choke on their own accounting at those levels. To look at the extremely-hypothetical router above, it will be easier to attach some digital bearer microcertificates to an information packet, so that the packet pays its way through all the routers it needs to go to, than it will be for some giant book entry system to account for it all. People have said that those microcertificates could work like stamps, where the first router cancels the stamp and pays back the other routers in the route some fraction of the "stamp" price to be settled later, or it may be possible to simply endow a packet with all the certificates nessary to get from point to point in a network someday. It kind of reminds me digital cash-as-processor-food, a bit.
Note that this kind of bio-economic thinking is not new, the Agorics folks and Stewart Brand have been talking about this stuff for quite a while. My point here is that "down" the network, not up, is the place to look for the interesting stuff in the future.
There are several interesting micropayment certificate systems out there, and there will be more. As software gets smaller and smaller with component architectures like OpenDoc and its eventual successors, it will be more and more economic to charge rediculously smaller and smaller amounts for rediculously smaller and smaller network behaviors. Just wait until someone figures how to get software to really "evolve", or gets software to write other software on a practical basis.
Most of the people I read, on the net or off, don't see this. They're looking "up" the net, how connections are made up, at the level of the grosser network features, like how monolithic corporations, or book-entry database and financial control systems, or government regulations, will happen on the net. How the net will integrate itself with the "real" world they're familiar with.
8 years ago, I used to talk about people who lived "on" microcomputers versus the ones who lived "in" them. I used to say that Macs were more for people who lived "in" computers because they weren't hindered by the mechanics of the interface so much. I think that there are still a lot of mainframe-cum-client/server folks out there who still live "on" the net, and not in it. Those are the people who are looking "up" at how the "big" players will behave, when they should be look at their feet, where the real action is.
Where the very ground is in the process of dissoving out from under them.
Cheers,
Bob Hettinga
[ Back to the previous Rant ] | [