Munching on munchkins.

Adam Rifkin (
Mon, 24 Nov 1997 18:23:56 -0800

Ron Resnick writes:
> Just in case anyone is keeping track, this is the last FoRK ever
> heard of I Find Karma. Obviously the guys at UFS got him for
> copyright violation, and he's now doing 5-10 years of hard time for
> Snoopy mockery.

Damn. I almost lasted two months without posting. Hope you enjoyed the
silence. Reports of my death have been greatly exaggerated. Actually,
I met the other me:

but that isn't why I broke my silence. Actually, it was

that broke me. Actually, it broke my jaw as it dropped to the ground.
And the fact that I wanted to complain about forkmembers not buying
enough books from Amazon.Com with the forkrecommendedrA attached to the
end as prescribed by...

Which reminds me, The Simpsons "A Complete Guide to Our Family" is the
most incredibly bitful book about any piece of entertainment ever
written. Only $12 at Barnes and Noble through December 11, or you can
of course buy it from Amazon and give a little something back to the
FoRK community:

Gosh, I should have a bit dump post where I extol the virtues of
beautiful new pieces of art like Metallica's "Reload" and Denis Leary's
"Lock n Load" -- two great new albums with "load" in the title... hmmm,

Or maybe what made me break the silence was the OMG-DARPA Workshop on
Compositional Software Architectures being held in Marriott Hotel,
Monterey, California, January 6-8, 1998. Doesn't looking at

just make you want to salivate? Made me drool out 3 pages worth...

Of course, I already know at least one position paper I like better...

But I digress. I want to go back to

> Defense Advanced Research Projects Agency (DARPA), Contracts Management
> Office (CMO), 3701 N. Fairfax Dr., Arlington, VA 22203-1714
> Douglas Dyer, DARPA/ISO, fax (703) 696-2201 E-MAIL: Technical and
> contractual questions, . AGENT-BASED SYSTEMS (ABS)
> BAA98-01 DUE 010798 POC Major Douglas Dyer, (Technical) DARPA/ISO, FAX:
> (703) 696-2201, WEB: (search under solicitation),
> E-MAIL: . The Defense Advanced Research Projects
> Agency (DARPA) Information Systems Office is soliciting proposals for
> multi-agent systems research and accompanying new technology
> development. Current military command and control software is manual,
> cumbersome, and exacting. During a crisis, military commanders and
> their staff waste precious hours manipulating information systems,
> manually specifying details, and performing simple tasks in order to
> assess the situation, form and execute plans, and protect the
> information infrastructure. Technology based on teams of intelligent
> agents acting autonomously to perform delegated tasks on behalf of the
> user offers the potential for easing, automating, and speeding many of
> these jobs. Intelligent agents must adapt to different problem solving
> situations and should be particularly powerful when they communicate
> and cooperate with other agents. Ensembles of cooperating agents will
> allow tailored sets of skills to be brought to tasks in a dynamic ...

Wow. Drool, drool. Sounds like they *need* munchkins. Due 1/7/98, eh?

Hmmm. Tap tap tap, do a little research...

> Deborah Estrin is currently an Associate Professor of Computer Science
> at the University of Southern California in Los Angeles where she joined
> the faculty in 1986. She received her Ph.D. (1985) in Computer Science
> from the Massachusetts Institute of Technology, her M.S. (1982) from
> M.I.T. and her B.S. (1980) from U.C. Berkeley. In 1987, Estrin received
> the National Science Foundation, Presidential Young Investigator Award
> for her research in network interconnection and security. Her research
> focuses on the design of network and routing protocols for very large,
> global, networks. Current research projects are in scalable multicast
> routing and transport protocols, self-configuring protocol mechanisms
> for scalability and robustness, and tools and methods for designing and
> studying large scale networks.
> Estrin is a co-PI On the VINT (Virtual Internet Testbed), NSF Routing
> Arbiter, and RSVP-II projects at ISI. She is an active participant in
> the Inter-Domain Multicast Routing WG and End-to-end research group.
> Estrin is a member of the ACM, IEEE, and AAAS. She has served on several
> panels for the NSF, National Academy of Sciences/CSTB, DARPA, and Office
> of Technology Assessment. Estrin is currently an editor for the ACM/IEEE
> Transactions on Networks. Estrin is a member of DARPA's ISAT advisory
> board, and a member of the board of directors of Precept Corporation.

Yup, munchkins are important. How can we convince her?

Let's look back at the last things on this list thought or said about
munchkins. For no other reason than to see if the above fits in with
the below, and makes sense economically. Sounds as important as
Tenenhouse's Active Networks leadership.

In the munchkin world, regardless of who invents munchkins or how, each
new munchkin in the world is a new asset, a new printing press of
money. It's analagous to a new human being: it is a self-regenerating
asset with value far out of whack with its creation cost. Each munchkin
is born with a NEW surplus supply of an initially valueless commodity:
its air time. Traditional economics is about scarce resources, not
infintely expandable ones. And modulo some basic physical limits, munch
bandwidth IS infintie. Cell sizes just keep decreasing; spectrum reuse
keeps increasing; fiber reevolves, etc.


In this sense, we're edging up on a FUNDAMENTAL economic question many
paths lead to: microcurrencies of all kinds -- all attention economies
-- lead to this question.

However munchkins are invented, protected, or sold -- actually, as long
as they ARE, indeed sold rather than rented -- in which we must trust
the US antitrust laws in much the same way they slammed IBM for leasing
mainframes -- there will be N(munchkins) incompatible local currencies
around. And every new munchkin is a new printing press. How do us gnomes
control the supply?

Well, we have to presume that due to marketing we control yet another
prestigious currency: Roebucks. Assuming perfect management, Roebucks
are accepted as a standard store of bandwidth value -- but the supply of
Roebucks, in equilibrium CANNOT be fixed. You can't control your own
printing press of Roebucks, either. It is like playing the
second-derivative game of the Fed today. Roebuck supply HAS to float
with the absolute number of bits/second available on the planet.

Decentralized currency creation does NOT obviate the need for a central
reserve currency -- but there is NO POSSIBILITY of a gold standard to
externally measure the expansion of local currency supplies.

In practice, our agents have to sit around the world and operate as
independent local exchanges, setting the prices of Roebucks per local
conditions -- but enforce UNIVERSAL value measures. That is, there
should NOT be arbitrage opprtunites in local roebuck spot markets (we
absorbed all of them already). It's a little like letting stockbrokers
run your monetary policy.

Now, Bits/second, as a currency, increases with every new munchkin on
the planet. EXCEPT, tower-of-babel like, there is no automatic
convertibilty of bandwidth claims. Three atoms of gold might buy 36 baud
from and 100 baud from my neighbor. And each new munchkin adds to the
'money supply' of bits/second.

To have a viable planetary telecommunication system, there must be a
reliable store of value that allow local bandwidth swaps (for example,
packet transit rights) to be convertible across days, months, years, and
across locations.

The world economy of bits, thus, has been born as a myriad of tiny local
exchanges. 'Local market area' has meaning again! -- what irony! Of
course, within my own physical territory, I completely control prices --
it's an internal currency for household or campus bitflows. Across
borders, I can only make meaningful promises for exchanging future
capacity with my local neighbors.

YET-- we believe in the interconnectedness of all locals. There is a
single monstrous simultaneous equation being solved here. We can
hypothetically compute the price of a universal currency from the
residual data -- but can we actually create that currency, reify that
stable point moment-to-moment as an actual product: Roebucks (TM)?

Worst case is that the exchange currency normally evolves out of the
chaos on its own -- then no one makes any money on it! :) :) :)

We expect that because some users have REAL needs for bandwidth
COMMITMENTS, there will be a MARKETING niche for a new currency.
Roebucks will have to be backed by an insurance policy -- we guarantee
that bit transactions will clear at the stated price even if we have to
use our own equipment to reroute over wirelines or wildly expensive
Teledesic links. This is the same INSITUTIONAL commitment as a stock
exchange has.

So Roebucks, a premium product, will be priced ABOVE the mythical
stable-point exchange currency. We will make money on the arbitrage --
the value of the insurance of our brand name.

The consequent goal is to aim for market SHARE -- to persuade more
contracts to be cleared in Roebuck-denominated terms than any other.
Clearly, this only applies to the fraction of bit-barters that occur
ACROSS trust boundaries.

Intra-hong-kong transactions can be freely denominated in kongbucks --
but we all know that if a deal REALLY matters, it is done in
greenbacks. Greenbacks have a brand-name premium. Only the british pound
and the deutchmark and perhaps the yen might make similar claims. What
we need to do, what has not been done before in the history of central
banking, is to privatize the branding of currencies and translate that
prestige into profits.

Will a micro-example help? Suppose I buy two munchkins, and program them
to trust each other. One is at my apartment, one is at the lab. They can
set up a line-of-sight connection. Within this link, bits are auctioned
off between email, video-mail, and quake sessions in terms of some
completely private munchkin-specific currency: Rifferbucks, say. Now, if
Caltech installs a munchkin network along ring road, I have a problem. I
have to also bid for the airtime in contention with official Caltech
transponders. In the end, they own the land, so they should also own the
spectrum rights across it, and I am screwed. They could even go so far
as to require all airtime to be paid for in zotbucks, so I have to trade
some real-world currency for the right to continue retransmitting.

In fact, though, I have a valuable asset to barter with. My munchkin
pair COULD add to the total supply of baud for sale on campus. So if I
relay enough zotpackets, I really should be able to earn enough to 'pay'
for my own transmission needs. I'm still at Caltech's mercy for the
exchange rate. They may say I have to relay ten packets to send one: a
10:1 zotbuck to Rifferbucks ratio.

Free markets are the proper response to state-controlled allocational

Suppose MCI, which has the Caltech payphone contract, wants to use
munchponders to run their payphone network on campus. NOW it gets
interesting. MCI has links to the EXISTING datasphere: fiber and
microwave links to the existing Internet. Caltech is salivating over
trying to use its munchnet to move Internet traffic as well as Intranet
traffic. MCI can now treat Caltech like an end-terminal say that well,
we'll take one of your zot-packets around the world for every five you
allow us to transmit locally. So now MCIbucks are at 5:1 with
Zotbucks. Caltech might be miffed at the price, and may even decline to
clear any barters at this expensive rate. We can buy traditional
Internet bandwidth from Teledesic at x dollars per baud, and frankly, we
think (pull number out of ass) that's cheaper than our wireless
zotpackets (cost of network / total baud density of campus).

Two things happen. As the number of zot munch nodes goes up, spectrum
reuse and baud density goes up. The 'price' of zotbucks goes down enough
that they may take MCI's 5:1 over continuing to pay Teledesic (or some
tradeoff function thereof). Second, MCI may treat Caltech as a transit
node, not just a bitsink. If they install payphones in Old Town Pasadena
Marketplace AND in Downtown LA, zotpackets become valuable to them for
transit. So they could change their offer to 3:1.

Multiply this by MCI negotiating with millions of land-owners, and MCI
bucks slowly gain currency <pun> as the gold standard <irony>. Now, MY
transmitters get back in the game. Sometime during these market
fluctuations, it may have been advantageous for me to exploit the
floating rates between Rifferbucks and zotbucks and zotbucks to mcibucks
to directly arbitrage packets from Rifferbucks to mcibucks. This is a
*nanomarket*. But suppose my fiancee, the only person within range who
might realistically choose between using Rifkin munchkins and Zot
munchkins to reach the outside MCI net, can help me create a viable
COMPETING exchange market in Rifferbucks:zotbucks alternative to the
campus's artifically mandated 10:1.

Such disputes will eventually be appealed back up the chain of
conversion transactions to a single, commonly trusted reserve
currency. Initially, when the ultimate competitor was teledesic, it
would be dollars, since that's what Craig and Bill want. If MCIbucks are
reliable (and plentiful enough for peak demand -- like a run on a bank),
then we can both denominate our service in MCIbucks. And if MCI is
really, really, on the ball, well, we could just decide to abolish
xentimes and zotbucks in exchange for a BILATERAL agreement where we can
just mint our own MCIbucks -- we sign over the branding agreement so
that what our munchkins, whether owned by Rohit or the Univesity of
California -- simply denominate all cross-domain transactions in
MCIbucks. In exchange, MCI gives up epsilon control of the currency,
since the global value of MCIbucks (how much baud & QoS you get in
exchange) is partially at the mercy of the infrastructure now owned by
others. MCI earns its 'master franchising fee' by insuring that whatever
bonehead mistakes happen to UCI's physical infrastructure or mine, they
will route Bank of America's traffic from Irvine to Newport by way of
fiberline if need be.

And if their advertising geniuses can continue to portray MCIbucks as
the friendlier alternative to Vaderbucks from New Jersey, well, MCI
stands to make a big profit. Because, somewhere, some schmoe just
realized he ABSOLUTELY, POSITIVELY needs a videoconfernce from Omaha to
Pasadena and never bothered to play into the communal pot by stockpiling
his own packet chits by selling transit rights. And MCI will sell from
its reserves in exchange for DOLLARS -- converting the ethereal exchange
value of surpulus baud capacity into real-world goods and services.

Or maybe not -- that schmoe might want to barter with some Buffetbucks
good for accessing his stock-picking newsletter. And that might be the
death of the greenback.

EPILOGUE. The question of how a reserve currency gets its hump is a
fundamentally open one. Any theory that can actually predict the value
of reserve premia out of the parameters describing the volume and
fragmentation of local markets could well unravel the mystery of the
dollar's power -- and net a nobel prize. Like most economics nobels,
it's probably a subtle consequnce of some basic operations, in this
case, the future market for bandwidth and the insurance costs of
'bridging' fragmented local markets in crisis (by analogy, the cost of
actually moving productive capacity amongst national currencies -- like
shipping wheat). This is an essential and compelling question because
the decreased friction of the net WILL create a boom in
cooperatively-valued currencies outside the control of any central bank
-- at worst, we'll have to wait for Chaum's patents to expire. The
economic theory to return control appears to be utterly unexplored.

Sound familiar? Just seeing if anyone read Rohit's letter to me a month


That's how you're gonna beat em, Butch. They keep underestimating you.
-- Pulp Fiction