IP: Another not competely different view -- A response to Faulhaber

[David Farber <dave () farber net>]

Gerry is well aware of some of the options enabled by modern technology -- 
like SDR etc. We he and I , as well as Dave Reed believe that rethinking is 
in order and all start with a position which will, from past experience, 
change as the debate unfolds.

Dave

Co-Director The Penn Initiative for Markets, Technology and Policy

> Date: Mon, 20 Aug 2001 12:07:46 -0400
> To: farber () cis upenn edu, "Gerald Faulhaber" <faulhaber () wharton upenn edu>
> From: "David P. Reed" <dpreed () reed com>
> Subject: A response to Faulhaber
>
> Dave - feel free to send to IP if you want, but I think it is important to 
> take issue with a key aspect of Prof. Faulhaber's comments about the 
> utility of a market in allocating spectrum.  While I agree that the old 
> (GOSPLAN?) model doesn't work well, to go to a "spectrum as property" 
> model quickly could be even worse, because it is based on a *very* 
> seriously flawed metaphor.  We can and must do much better than this.
>
> Let me preface this comment with a caveat:  I am a very strong advocate of 
> markets as allocators of resources.  Where such markets can be made to 
> work, they are simple and incredibly effective.  We should use them much 
> more than we do - where they apply.
>
> But the electromagetic fields (whether used for communications services, 
> information distribution services, or location services like GPS, radar, 
> or LORAN) do not have the physical properties that can properly be managed 
> by a market of the sort that Faulhaber advocates.
>
> Rather than a legal system based on a faulty understanding of the physics 
> and  network properties of communications, we need a much better 
> approach.  We need one that looks forward to the deployment of 
> technologies that are only in their early states, rather than one that 
> looks backwards to a metaphor based on the primitive, narrow-band, 
> point-to-point radio techniques known in the early 20th century.
>
> What is essential to a property market is that the resource (property) 
> being allocated be:
>
> 1) naturally subdivided or subdividable without appreciable loss, into 
> individually transferrable units,
>
> 2) relatively expensive to create or produce,
>
> 3) where the transferrable units are can be used independently without 
> impinging on the value of other units.
>
> Land, gold, etc. fit these criteria reasonably well.
>
> Electromagnetic fields (excited by antennae) do not.
>
> Historically, the US (with other countries following close behind) came up 
> with its "spectrum policy" in a hurry, after the Titanic disaster showed 
> that there was a public interest in making sure that radios were operated 
> in the interest of public safety and public values.
>
> But the structure being used today is very much out of date, and does not 
> conform to what we understand about the properties of communications 
> through our physics and our sciences of information (information theory, 
> communications theory and computing sciences).  Though the old licensing 
> model for spectrum allocates bands in ways that superficially resemble the 
> subdivision of property among users, this is not well-matched to the 
> physics or technology of modern communications systems, especially those 
> that include networking and computational elements.
>
> There are three serious problems with trying make the metaphor of 
> "spectrum" as "property" work.
>
> 1) SUBDIVISION IS WASTEFUL. Dividing spectrum up by "frequency" is done by 
> inserting an arbitrary guard-band between users, along the dimensions of 
> frequency (the narrower-band the signal, the larger the proportion of its 
> band allocated to unused "guard-band"), power (the wider the variation of 
> power among stations in a band, the more low-power stations are displaced) 
> and space (most of the geographic area allocated to stations sharing a 
> frequency is in the areas of weakest signal at the periphery of their 
> range.  This means that subdivision is NOT without loss, and in fact the 
> more subdivided the spectrum is *by the current policy approaches* the 
> more waste of capacity there is.
>
> Simple approaches discovered relatively recently, like OFDMA (orthogonal 
> frequency division multiple access), recognize that by using a better 
> *architecture* that requires serious cooperation among e-m field users, 
> much of this waste can be avoided.  But OFDMA is an architecture for 
> sharing that does not subdivide spectrum into independently marketed 
> pieces.  Instead it creates a common communications channel that can be 
> allocated in real time among users, giving them much more effective 
> utilization.
>
> 2) NETWORKING CREATES VALUE, PARTIONING DESTROYS VALUE. Network value can 
> scale according to an increasing returns model: as more stations 
> interoperate, the value available to all increases faster than the number 
> of stations.  The scaling laws now called "Metcalfe's Law" and "Reed's 
> Law" are examples of the increasing returns enabled by network 
> architectures.  The main result of increasing returns is that there is a 
> huge cost to "balkanizing" networks - reducing their interoperability. For 
> example, Metcalfe's Law would have clarified the societal cost of the 
> business decision by Telefunken and Marconi in the days of the Titanic - 
> Marconi operators were not allowed to communicate with Telefunken ones, 
> and vice versa.  By dividing the world in half, the total "value" 
> according to Metcalfe's Law was 1/4 as large.  Dividing interoperability 
> into 8 independent operators would similarly divide the value available to 
> stations by 64.  Thus, even early regulation required - at least for 
> public safety - that the networks be interoperable, and carry emergency 
> traffic for each other's customers.
>
> Today, we know how to make networks of all types interoperate.  It's that 
> magical virtual "ubernetwork" we call the Internet.  Yet a mere 
> subdivision of spectrum, auctioned of to non-interoperable communities is 
> hardly likely to result in interoperability - instead it seems that the 
> more likely result is a collection of tiny walled gardens with 
> t(r)ollbridges between them, vastly reducing or eliminating interoperability.
>
> (footnote: theoretically, we can also make location-service networks 
> interoperate as well - so we don't depend on flaky systems like the 
> marvelous-where-it-works GPS. The "Interlocate" architecture is begging to 
> be developed IMHO, if we don't auction off all the spectrum too soon).
>
> And this will vastly undervalue the resources available.
>
> 3) PROPERTY RIGHTS INHIBIT INNOVATION THAT WOULD IMPLY DIFFERENT 
> STRUCTURES. Innovation in physical communications has only recently begun 
> to yield technologies that a) don't fit the "narrowband" fixed spectrum 
> allocation well at all, and b) provide vastly more effective and efficient 
> use of spectrum for new applications.  Here I'm talking about adaptive 
> software-defined radio, cooperative joint detection, adaptive ad-hoc 
> networks, ultra-wideband (time-domain) modulation, etc.  What's common 
> about all of these technologies is that they are inherently co-operative 
> systems that share a common electromagnetic field among many users, 
> dynamically cooperating to share that field more efficiently than any 
> static allocation makes possible.  Combintations of these approaches are 
> likely to provide systems that give total systems capacity that grow 
> perhaps as fast as linearly in the density of stations.  Traditional fixed 
> spectrum allocation provides a system capacity that does *not* grow with 
> station density - it stays flat at best, and more commonly declines.
>
> Taken together these three observations, which arise only in situations 
> where spectrum is shared under an architecture that does not balkanize 
> spectrum, mean that the market system Faulhaber proposes risks
>
>         a) blocking valuable innovations,
>
>         b) suboptimizing the allocation of a public resource, and
>
>         c) creating a set of unnecessary "bottlenecks" that create wealth 
> while
>            degrading the potential value that could be available to the 
> public
>            under a different economic allocation.
>
> For an example of (c) - a real-time market in "information capacity" can 
> be built where stations cooperate in real time to allocate capacity using 
> software defined radios and internetworking.  This can be run as a 
> "market", but since the units being traded would be "messages", they could 
> be priced directly in terms of their utility.  And if the systems 
> architecture can produce much more capacity for "messages" by new 
> technology and cooperation rather than balkanization, the overall utility 
> will rise in proportion to the cooperation exhibited.  Thus the incentives 
> would be aligned to allocate capacity towards its real value, and more 
> importantly to incent the creation of new value using cheap means such as 
> internetworking.  This would combat the alternative strategy sought by 
> owners of mis-designed property rights - that of using their monopoly to 
> maximize returns while blocking effective arbitrage/bypass strategies that 
> would lower the value of their property.
>
> The creation of a sensible economic structure for these new radio 
> architectures based on networking and computation will be a 
> challenge.  Prof. Faulhaber is clearly aware of some of the potential, but 
> I'm afraid that the property rights auction approach is *not* the only 
> economic model that is available, and in my opinion it is far from the 
> right one.
>
>
> - David
> --------------------------------------------
> WWW Page: http://www.reed.com/dpr.html



For archives see: http://www.interesting-people.org/