\documentclass[runningheads,12pt]{llncsx2}
%In order to omit page numbers and running heads
%please use the following line instead of the first command line:
%\documentclass{llncs}.
%Furthermore change the line \pagestyle{headings} to
%\pagestyle{empty}.

% \usepackage[]{graphicx}

\input{psfig.sty}

\def\tm{\raise5pt\hbox{{\rm\tiny TM~}}}

\begin{document}

\pagestyle{headings}
%In order to omit page numbers and running heads
%please change this line to
%\pagestyle{empty}
%and change the first command line too, see above.

\mainmatter

\title{Pricing and Architecture of the Internet: Historical Perspectives
from Telecommunications and Transportation}

\titlerunning{Pricing and Architecture of the Internet}

\author{Andrew Odlyzko}

\authorrunning{Andrew Odlyzko}

\institute{Digital Technology Center, University of Minnesota\\
499 Walter Library, 117 Pleasant St. SE\\
Minneapolis, MN 55455, USA\\
\email{odlyzko@umn.edu}\\
% \texttt{http://www.dtc.umn.edu/$\sim$odlyzko}}
\texttt{http://www.dtc.umn.edu/$\sim$odlyzko}\\
\texttt{Revised version, August 29, 2004}}


\maketitle

\begin{abstract}

With telecommunications in a slump,
the search is on for ways to re-invigorate this key industry.
The main problems are clearly economic much more
than technological, and many of the proposed
remedies would lead to new architectures for the
Internet that would provide for greater control by carriers.
They would drastically reduce the role of the
end-to-end principle, the main foundation for the
success of the Internet, in which
functionality resides at the edges of the network.
The proposals to restrict voice over Internet (VoIP) are
just one part of this trend.

\vspace*{+.05in}

~~~~Historical precedents from telecommunications for introduction of
differentiated services and sophisticated charging
methods on the Internet are discouraging.  The
general trend 
has been towards decreasing
price discrimination and simpler pricing.

\vspace*{+.05in}

~~~~The history of transportation presents a different
picture, with frequent movements towards increasing
price discrimination and more complicated pricing
(although with many noteworthy reversals).  
Charging according to the nature of the goods
being transported has been and continues to
be the norm.  Since
the incentives to price discriminate are
increasing, and the ability to do so is also
growing, it is conceivable that telecommunications
might break with its historical record and follow
the example of transportation.  It is therefore 
of interest to examine the evolution of pricing
and quality differentiation in transportation.

\vspace*{+.05in}

~~~~Some historical sketches on the evolution of
pricing in transportation are presented.  Their
implications for telecommunications, and 
especially for Internet pricing and architecture,
are discussed.



\end{abstract}


\section{Introduction}

The telecommunications industry at the end of 2003 is
perceived to be in a deep slump.  A closer examination
shows that in spite of many bankruptcies,
total revenues of the telecom service sector as a whole are
reasonably healthy, having suffered at most a slight decline,
following a period of very rapid growth
\cite{Odlyzko17,Standage}.  On the other hand, the telecom
supplier sector as well as stock market valuations of almost
all telecom enterprises have collapsed, as has the aggregate
investment level.  This reflects a collapse of expectations
that telecom revenues would skyrocket.

There is a concern about the ability of
the telecom industry to deploy innovative new services that 
the recovery of the information technology (IT) industry
as a whole may depend on.
The problem is that not only is this industry dealing with
gross overcapacity in many sectors, but it may not be
structured properly for the future.  Technology is not
the main problem, as it is more
than adequate for the demands placed on it, and is still
advancing rapidly.  It is the economics of future telecom services
that is in question.  In a memorable phrase attributed to
David Clark, ``we know how to route packets, what we don't
know how to do is route dollars.''

In searching for a way out of its slough of despond,
the telecommunications industry is considering major
changes in the architectures of networks, changes that
go counter to recent trends.  As an example, in early 2003, a few months
after AT\&T pioneered 
a flat-rate domestic long distance calling
plan (after resisting valiantly for many years, along with
the rest of the industry), followed later by an international
flat-rate plan, a high-level
manager from AT\&T as well as a representative of
another service provider talked at an industry meeting about the
critical need to bring usage-sensitive pricing to the Internet.
Since that time, the flat-rate bandwagon has continued to
pick up speed.
Sprint started offering in August 2003 a
flat-rate plan that includes all types of domestic voice calls, wireless as
well as wireline.  On the other hand, industry leaders
continue calling for new network architectures that
would include as a key feature measured rates and have
application-aware cores \cite{Bradner}.
These desires by industry leaders for new architectures
are not just empty dreams.  They also affect their
day-to-day actions and the evolution of the current
network.  As one example, peering of VoIP traffic
is not progressing much at traditional
service providers.  The main obstacles are not technical,
but economic ones, as carriers are waiting to establish
new business models.  The industry's desires are reflected
in the words of one official
carrier representative, who predicted
that ``[VoIP peering] won't be handled like 
Internet peering, where we all terminate each other's
traffic without paying fees; it's going to follow the voice model'' 
\cite{Maitland}.  (In the meantime, independent VoIP service
providers are happily growing their business through offerings which manage
to ride over the ordinary Internet with its standard peering
arrangements.)

How do we reconcile this contrast between what is happening
in the marketplace and what is desired by the service providers?
The telecom industry could simply be irrational in its
attitudes.  There has been a trend, going back centuries,
of preoccupation with content and undervaluation of connectivity \cite{Odlyzko8},
and a persistent lack of understanding of the advantages
of simple pricing \cite{Odlyzko7,Odlyzko10}.  Many times
the industry was forced against its will into simple pricing or else accidentally
stumbled into it, and yet eventually this worked out
for the good of the service providers.  However,
the industry views simple pricing as the disease
that has contributed to the current telecom depression.
The wireless industry, in particular, has often boasted
that it managed to avoid the mistakes of the Internet
by avoiding the
open architecture and flat-rate pricing of the latter. 
And indeed, cellular revenues continue to grow, while
those of the rest of the industry decline or at best
stay constant.  The spate of
bankruptcies and downsizings that we have witnessed in
telecom was most pronounced in the Internet sector,
and has largely bypassed wireless.  (Of course, there
have been giant duds in the wireless arena, as with the 3G spectrum auctions.
There have also been
major disasters in related areas, in particular
in low orbit satellite communications, with Iridium and
competitors all going bust.  But on the whole, the wireless
industry has managed to absorb such disasters without the
extreme pain experienced elsewhere in telecommunications.)
Cellular today serves more people and has higher revenues
than the Internet, both in the U.S. and (even more so)
in the rest of the world.  Thus it is understandable
that the rest of the telecom industry might like to
imitate the wireless sector.

Historical precedents in telecommunications argue
for continuation of the trend towards
simplicity.  (See \cite{Odlyzko10}.
Section 7 has a brief survey.  Note
also the early work, dating back to the late 1980s,
of Anania and Solomon \cite{AnaniaS},
predicting that flat rates would dominate for data
networks.)
There have also long been arguments based on distribution of costs of 
the entire telecom sector that suggest simplicity will
minimize total system costs, even though it might lead
to seemingly inefficient use of transmission and switching
\cite{Odlyzko2,Odlyzko5,Odlyzko6}.  
These arguments are even more persuasive now, when
transmission prices have crashed, then in the mid to late-1990s,
when those prices were rising \cite{CoffmanO1}.
The counterargument,
though, is that this trend has been shown by the telecom crash to lead to ruin.
If service providers can only sell 
commodity undifferentiated transport, and most of the
intelligence is at the edges of the network, in end-user
equipment, carriers will simply not invest enough
to build out the necessary networks.  And indeed, 
the cash flow of the telecom service industry is very healthy,
and debts have been whittled down.  On the other hand,
capital investment
is extremely depressed, as carriers wait for profitable
opportunities.

Thus even if it is not optimal from a global point
of view, it might be necessary to introduce complexity
in order to be able to construct and operate the
telecom infrastructure, especially the residential
broadband networks that are so eagerly awaited by
government and industry leaders \cite{Odlyzko17}.  
That might mean allowing
carriers to charge differently for movie downloads
than for Web surfing.  That, in turn, might require a new
network architecture.  Such a move would not be
unprecedented.  
The key (although seldom mentioned) factor behind
the push for new network
architectures appears to be the incentive to price discriminate.
It is an incentive that has been operating since the
beginnings of commerce.
Throughout
history we observe many practices, such as
early postal rates that depended strongly on distance and
weight \cite{Odlyzko7}, that can only be justified
in terms of need to price discriminate.
As yet another example, consider the following incident
from the early days of the telephone (p. 102 in \cite{Marvin}):
\begin{quote}
In Britain in 1889, postal officials reprimanded a Leicester
subscriber for using his phone to notify the fire brigade
of a nearby conflagration.  The fire was not on his premises,
and his contract directed him to confine his telephone ``to
his own business and private affairs.''  The Leicester Town
Council, Chamber of Commerce, and Trade Protection Society
all appealed to the postmaster-general, who ruled that the
use of the telephone to convey intelligence of fires and
riots would be permitted thenceforth.
\end{quote}
Such practices are motivated by the
wide dispersion in willingness to pay.  Different 
people are able or willing to pay wildly differing
amounts in total for all telecom services.  Further, individuals,
even with the same total willingness to pay, vary
substantially in how much they are willing to pay for
particular services.
As a simple example, Table 1 shows estimates of what
U.S. residential users pay per megabyte of traffic from various
sources.  The methodology used in constructing the table
is only approximate (for example, 
the cable TV rate is computed by assuming average daily
viewing time of four hours, with a signal carrying 6 Mb/s,
and the SMS rate comes from assuming an average of 33 bytes
per message, at 10 cents for each message), but the wide range of
prices is striking and persists even if we modify the
assumptions underlying the calculations within reasonable bounds.


\begin{table}[tb]
\begin{center}
Table 1.  Value of bits: Cost per megabyte of various services. \\
~ \\
\begin{tabular}{lrr}
service & typical &  revenue \\
        & monthly bill & per MB \\   \hline
cable & \$40 & \$0.00012 \\
broadband Internet & \$50 & \$0.025 \\
phone & \$70 &  \$0.08 \\
dial Internet & \$20 &   \$0.33 \\
cell phone   & \$50 &   \$3.50 \\
SMS &   &   \$3000
\end{tabular}
\end{center}
\end{table}

That business leaders understand the value of price discrimination
(even though they seldom talk about it publicly), is shown by an
email of August 17, 1997 from Warren Buffett, the famous investor and 
head of Berkshire Hathaway, to Jeff Raikes of Microsoft (made public
through a lawsuit, and quoted partially in \cite{WilkeC}):
\begin{quote}
[Alexander Graham] Bell should have anticipated Bill [Gates] and let
someone else put in the phone infrastructure while he collected by
the minute and distance (and even importance of the call if he could
have figured a wait [sic] to monitor it) in perpetuity.
\end{quote}

The basic problem that the telecommunications industry
is struggling
with is that of moving away from vertically integrated 
service providers, each with its own technology.
The natural evolution is towards 
a heterogeneous mix of physical access methods, which
will be largely invisible to the users, and will be
unified
by the Internet Protocol.  If this scenario plays out, the
end-user will have available a simple data pipe (of bandwidth
depending on location, willingness to pay, and so on), and
services such as voice or TV will be delivered inexpensively
over that pipe.  This will replace the current mix of
different vertically integrated networks designed for
different services.  The problem for carriers with this scenario is
that it is the basic transport service that is likely to
be the most expensive.  Services (at least the familiar
services such as voice) appear likely to cost
less and less to deliver (aside from fees to content
providers).

If one thinks of services in terms of bits (as much of the
industry and even research community do, even though there
is plenty of evidence this is misleading), then the great
disparity in valuations in the table above suggests keeping
different services separate, as SMS is kept separate in
wireless communication from voice.  
% Voice over IP (VoIP)
% is perceived as a serious threat to the telephone
% network since it threatens to divert voice traffic
% (which continues to provide the overwhelming bulk
% of telecom revenues) to lower-priced broadband services.
Even if one runs a single unified network, those disparate
valuations suggest that one should charge according to
the nature of the traffic, and not by the byte.  But that
goes counter to the nature of the current Internet.

% example of Beverley Beck, charters of 1726 & 1744
% per ton tolls, old & additional:
% salt in bulk  4  2
% iron or lead 4 8
% timber or stone 4 2
% hemp, line and flax 0 7
% sand 2 0

% also, flour is not listed in 1726, charged 0.75 per cwt (112 pounds)
%      in 1744

% but also had charges of 5 in old, and additional of 12 "for
% every other sort of goods, wares, merchandise or ladings whatsoever"




As I will show in this paper, price discrimination
had been practiced in telecommunications and transportation
for a long time.   
The value of differential pricing has been well understood
in economics since the middle of 19th century
(although the earliest research in this area has been
neglected in history books \cite{EkelundH}).  But it
has been practiced far longer.  Table 2 shows some of the tolls
charged by the Beverley Beck Navigation in 18th century
England.  (The full toll schedules were quite long, and can
be found in \cite{Priestley}, and will be mentioned later.)
The ``before 1744'' column shows the charges
according to the 1726 charter, the ``after 1744'' column
those after a revised charter was granted in 1744.  These
tolls were entirely separate from charges for carriage,
which were paid to boat owners.  The navigation project
did not have any costs that depended on the nature
of the cargo.  (The project's costs were primarily the 
costs of dredging the channel.  To the extent that
boats being towed through the channel created waves that damaged the banks,
or the towpaths got worn, the associated expenses
were the same whether the boat carried salt or iron or nothing
at all.)




\begin{table}[tb]
\begin{center}
Table 2.  Selected 18th century tolls on the Beverley Beck navigation (in pence per ton).\\
~ \\
\begin{tabular}{lrr}
cargo & before &  after \\
        & 1744 & 1744 \\   \hline
sand & 2 & 2 \\
timber, stone, salt & 4 & 6 \\
iron and lead & 4 & 12
\end{tabular}
\end{center}
\end{table}



Let us illustrate how this applies in the context of the
current controversy over VoIP.  The main advantage of VoIP is 
usually claimed to be greater efficiency in utilization of
network resources.  Sometimes this is explained as coming
from compression of voice signals and better multiplexing
of packets than of ordinary voice telephony streams.
Yet VoIP is arguably much less efficient in use of
basic transmission resources than traditional telephony.
Most current VoIP implementations do little if any
compression, in order to provide high quality.  
Moreover, because of the structure of
the industry, and the peering arrangements among Internet
backbones, most VoIP calls in the U.S. travel over extremely long
distances.  In contrast,
even ordinary long distance voice calls are mostly local, with
average distance of about 500 air miles between the two
ends \cite{CoffmanO1}.  Furthermore, most voice calls are local
\cite{Odlyzko7}, so the average distance of all voice calls
is on the order of 100 miles.  In additioin,
it was observed in 1998 that it then cost most corporations
more to send data over their internal long distance
data networks than it would have cost to send it using
a modem over the voice network, paying retail rates for
the voice calls \cite{Odlyzko2}.  Even today, that is
still probably true for many enterprises.
Analyses of current VoIP technologies for the enterprise
environment find few savings, and as a result relatively
slow adoption \cite{Bhagavath}.
Thus the efficiency argument for VoIP is questionable at best.
This argument was suspect from the very beginnings of VoIP,
since basic network costs of traditional voice telephony
were very low already in the mid-1990s, and systems for
compression of traditional voice telephony were widely
available (but were used primarily on links to less-developed
countries, and were used on a declining fraction of calls
even there).

In the long run VoIP is bound to win because of the new
features it makes possible, such as higher quality voice,
and integration with other services, as well as the advantages of not
having to run a separate network.  In the meantime, the
main incentive for VoIP comes from the ability to get
out of the elaborate maze of cross-subsidies, discriminatory
pricing policies, and taxes that are built into the
current telecom system.  The question is whether the
telecom industry can survive in the broadband era
without another maze of
cross-subsidies, discriminatory 
pricing policies, and taxes.  That the Internet has thus far
developed with an open architecture and simple
pricing does not mean that it can do so in the future.
The Internet has been a small factor in terms of revenues
so far, and has benefited tremendously from being
able to exploit the infrastructure built primarily
for voice services.
Currently spending on the Internet in the U.S. is only
around 10\% of total telecom revenues, as is discussed in Section 8.
The question is how the industry can evolve as all
the traditional revenue-producing services begin to ride
on top of the Internet.
To illustrate what the apparent difficulties are,
let us consider a simplistic economic model.  Suppose that a carrier
has two customers, A and B.  Suppose further that
A has a single phone line, and pays each month \$20
for the basic connection (including unlimited local calls),
\$30 for long distance calls, and \$20 for an Internet
account (with dial modem access, using the basic phone line).
Suppose that B, because of some combination of greater income 
and greater interest in telecommunications, pays \$40 per
month for two phone lines, one to be used primarily to
access the Internet, \$20 for an Internet account, and
\$70 for long distance calls.  
Thus the service provider
receives \$160 per month from A and B (if, as is
common, a separate ISP provides Internet connectivity).

Suppose the carrier now offers broadband connectivity.
The superior quality makes A willing to spend a total
of \$80 per month, and B a total of \$150 a month.
But suppose that they can get the services they
want (such as voice) from some other suppliers, on
top of their basic broadband pipe, for \$10 for A
and \$20 for B.  Then the willingness to spend on
that basic connection is \$70 per month for A and
\$130 for B.  If the carrier can get that much from
each, it can get \$200 per month, more than it used
to with traditional telephony.  That might 
make it worthwhile to provide the broadband
connection.  But if the carrier
can only offer a basic broadband connection for a
uniform price, the most it can get is \$140 per month,
which might not be sufficient to pay the costs of
providing that service.  That is the real dilemma
for telecom service providers.  Can they extract
enough money from their customers to pay for broadband,
if broadband is just a pipe?
Note that the problem can be solved if the carrier is
granted the power to price discriminate, and charge
customer A \$70 and customer B \$130.  

Giving carriers the power to price discriminate could
provide more funding for networks not only from customers,
but also from content providers.  Suppose there are many
residential customers for a carrier, and digital versions of
two movies, call them X and Y, each one of 2 GB, are offered by a studio.
Suppose that the studio can get \$5 for X per household
through DVD rentals or movie theater tickets, and \$10 for Y.
Suppose that we have a secure digital movie distribution
system (either through technological solutions or through
draconian enforcement of copyrights) so piracy is not a problem, and all customers
have the same tastes, and
are willing to pay an extra fee (compared to renting
a DVD or going to a movie theater) of \$3 for X and \$5 for Y.
The film studio and the carrier, both being profit 
maximizing entities, will ensure then that the price per movie
paid by consumers for home consumption is \$8 for X and \$15 for Y.
If the carrier gets to charge per byte, and charges \$1.50
per GB, it will collect \$6 per household for downloads
of both movies.  The studio will get \$17 per household.
If the carrier can control traffic on the network on a fine
scale, it can use differential pricing, charging \$3 for X
and \$5 for Y, for total revenues of \$8 per household.
In that case the studio gets only what it would from
other distribution channels, namely \$15.  (In practice, of course,
the studio would have to get some cut, but it might not
be very high.  Today, cable TV networks in the U.S. spend
only about a quarter of their revenues on content, the rest
going into their pockets to provide for network
construction and maintenance, overhead costs, and profits.)

As our economy evolves,
the incentives to price discriminate are increasing,
as fixed costs grow while marginal costs decline.
At the same time, the ability to price discriminate
is growing.  Decreasing privacy along with development
of digital rights management tools and the move away
from outright sales to licensing provide improved
methods for determining how much customers are willing
to pay, and to prevent arbitrage, in which those able
to buy at low price resell to those the producers insist
on charging high prices to.  (The prototypical example
is the positive passenger identification system,
which enables airlines to enforce their contracts, which
are for personal service to specific individuals, and are
not transferable.  See \cite{Odlyzko18} for more detailed
discussions.)

Charging by the value of the traffic (or some approximation
to it) runs counter to the basic architectural principles
of the Internet, principles that were key to its success.
The end-to-end principle \cite{SalzerRD} is the most
important of these, and says that functionality should be
be concentrated at the edges of the network to the maximal
extent possible.  In addition there is the open architecture,
and lack of discrimination, with all packets treated equally,
enabling end users to create new services without involving
carriers.  This leads to the image of the Internet
as the ``stupid network'' \cite{Isenberg}, which simply
takes packets and delivers them to their destinations.
(For further discussion of the effects of an open architecture
built on end-to-end principles on innovation in telecommunications,
see Section 6.)

What are the prospects of major architectural changes
in the Internet?  Traditional telecommunications regulatory 
literature does not offer much help, in spite of its
sophistication, with considerations of Ramsey prices
and the like, as it was developed for a relatively
static technology.  This paper presents some historical
sketches of the evolution of several network
industries in telecommunications and transportation.
One can question just how relevant these sketches are, since they
concern obsolete technologies.  Still, the basic
problem is not novel.  It is how to charge for a network service,
especially when the marginal costs of using it are
low.  Moreover, at the level of business models,
the approaches being proposed now are basically the
same as have been used before.  Thus it might be
interesting to observe how the questions we are facing
today were answered in the past.

As was mentioned before, in telecommunications,
the trend in general has been towards increasing
simplicity in pricing.  On the other hand, in transportation
we find many instances of increasing sophistication
in pricing, and almost a general principle of
charging according to the value of the goods being
moved (and thus without the end-to-end principle and
without privacy).  Thus those arguing for a new architecture
for the Internet that would limit its openness and
ability to innovate do have numerous historical
precedents on their side.

Many opponents of increased carrier control over telecommunications
are concerned about restrictions on political or personal activities
this might entail.  And there are certainly many governments that
are attempting (in many cases successfully) to restrict what
their citizens can do on the Internet.  The history of transportation
shows pretty clearly, though, that the main impetus for fine-scale
control over goods being transported came from economic incentives.
Charging tolls for canal transport of iron that were six times
higher than those of sand (see Table 2)
was a reflection of desire to maximize 
revenues, not political control.
This reinforces the surmise that economics is the main threat
to the current Internet architecture.

There is no definitive treatment of the history
of pricing in various network industries.  Even some
basic questions, such as the nature of the ``just price''
doctrine of medieval scholars, are still being debated.  
What is indisputable, though, is that price discrimination
has been a central concern for a long time.  Common law
and legislation have attempted to restrain it, but at
the same time to allow it to a substantial degree, in order not
to lose all benefits from differential pricing.  To cite
just one example, a British court decision of 1869
(cited on p. 191 of  \cite{Kostal}) declared that
``At common law, a person holding himself out as a common
carrier of goods was not under any obligation to treat all
customers equally. ...  All that the law required was that he should
not charge any more than was reasonable.''  

The next few sections (2 through 5) outline what happened
with lighthouses, canals, turnpikes, and railroads. 
There has been and continues to be a pervasive tendency
in these industries to price according to the value
of service, which requires information and control
over what was being carried, in complete opposition
to the spirit of the end-to-end principle.  
The general conclusion is that
there are extensive historical precedents
as well as arguments based on economics for new network
architectures
that would allow carriers increased control.
However, there are also strong arguments for
open networks.  
Then Section 6 discusses the connection between
open architectures and innovation, especially in telecommunications.
Section 7 is a brief survey of developments in 
telecommunications pricing.  Section 8 is devoted to
some projections of future evolution of the telecommunications
industry.  Final remarks are presented in Section 9.
Open networks not only allow for greater
innovation, but are more consistent with the
likely evolution of telecommunications, towards
a variety of technologies providing an interoperable
network unified by the Internet Protocol, and
with costs concentrated at the edges.  They are
also far more consistent with the nature of the
demand for telecommunications.  Hence it
appears that there are good prospects for the
preservation of an open architecture, although this
is not a foregone conclusion.






\section{Coase's lighthouse myth and other maritime tales}

Ronald Coase's 1974 paper \cite{Coase2}, ``The lighthouse in economics,'' 
has had tremendous impact on economic thought.  It claimed to show that
lighthouses, which had often been cited as prototypical 
examples of public goods that
only governments can provide, had in the past been provided in 
England by private enterprise.  This suggested to many that the role
of government could be shrunk, as many of its functions could be
provided by profit-making entities acting in their own interests, 
hopefully with some gain in
economic efficiency.  The Coase paper continues to be cited frequently
as a breakthrough result.  Unfortunately, many conclusions drawn from Coase's
paper are unjustified.
As an example, a recent commentary \cite{Warsh} claimed that Coase
had shown that ``[i]nstead of the government-sanctioned 'light dues'
charged by Trinity House, 
developers persuaded ship-owners to sign up in advance for voluntary tolls.''
This claim from \cite{Warsh} is incorrect, as are many
of the conclusions commonly drawn from Coase's paper.  In defense
of Coase, it has to be said that in his paper he never referred
to ``voluntary tolls'' and did not make the extreme claims
some of his followers have put forth.  However, Coase's paper uses very ambiguous
language to describe the English lighthouse system, and is deeply flawed.
Amazingly enough, even though any serious economic historian
should have been able to see instantly the faults in the paper,
it continued to be accepted uncritically for a quarter of a century.
The first debunking appears to have been by Richard
Epstein in 1999 \cite{Epstein}, and a more thorough one was
presented by Daniel Davis in 2002 \cite{Davis}.  (For an earlier, 
more general, criticism of the Coase paper, see \cite{Varian} which
has an excellent short summary of the key issues related to
public goods.)  These contributions are not
very well known even today, though.  

For an easily
accessible history of British lighthouses, see \cite{Trethewey},
from which much of the information below is taken.
What are the basic
problems with Coase's paper?  It is true that lighthouses
were frequently constructed, operated, and owned (with rights
of inheritance and sale) by private individuals.  However, they
were only exceedingly seldom the result of freely negotiated private contracts.  
They were usually the result of a grant (called a ``patent''
in those days) from the king, sometimes for a few decades,
sometimes in perpetuity.  This grant entitled the holder
to construct a lighthouse in a particular location and to
collect compulsory fees from all ships entering nearby
harbors.  The fees were set by the terms of the grant,
and were often collected by government customs agents.
Moreover, lighthouse owners did not just have the right
to collect the dues.  They had the obligation to provide lighthouse
services.  We read, for example, of King James I in 1623 reducing 
in half the compulsory levy for a lighthouse that was poorly
maintained \cite{Trethewey}.  There were instances of grants
that did not entail the right to collect compulsory lighthouse duties,
but those (as anyone before Coase would have predicted) were
generally not successful \cite{Trethewey}.

Some of the privately owned lighthouses were extremely
lucrative.  When they were taken over by the government
in the 1830s and early 1840s, owners were compensated.
The record payment was to the owners of
the Skerries Rock lighthouse, who received approximately
\$2.2 million.  (In comparison, the Louisiana Purchase
cost the U.S. \$15 million, and the acquisition of Alaska \$7.2 million.)
Still, these private lighthouses were clearly agents
of the government.  They had the duty imposed by the
government of providing services.  They also had the
power of compelling payment (at least for ships entering
British harbors), even from ships that did not rely on
them.  Shipowners did not have the option
of saying that since their ships would only sail in daylight
and fair weather, they did not need lighthouses and did not have to pay.  
The lighthouse arrangement was similar to many that
governments used to employ in order to compensate for
the lack of information and control technologies.
``Tax farming'' and even sales of officer positions in armies
were common in Europe well into the 19th century.
Yet it seems that everybody agrees that these were public goods
provided by governments.
Thus Coase's lighthouse paper certainly fails to support the thesis
that is ascribed to it in places such as \cite{Warsh}.
A good overall evaluation of Coase's paper is
in a phrase that Coase used to refer to the works of John Stuart Mill,
Pigou, and Paul Samuelson.  It makes
``statements about lighthouses which are misleading as to
the facts, whose meaning, if thought about in a concrete
fashion, is quite unclear, and which, to the extent that
they imply a policy conclusion, are very likely wrong'' \cite{Coase2}. 

The discussion of Coase's paper serves to point out how
misleading myths can survive for a long time, even when
there is abundant evidence showing they are false.  It also
leads to the heart of this
section, namely charging.  British lighthouse dues were
fixed by royal charters.
Lighthouse owners could not modify
them unilaterally to maximize their profits.
Still, there was evolution in charging.
What we observe is that as time went on,
schedules of fees tended to become more sophisticated.
The first recorded grant, from 1261, provided for a fee of
two pence for each ship \cite{Trethewey}.
By the 16th century, we see payment schedules of 6 pence for
a two-masted ship, 4 pence for a one-masted ship, and 2 pence
for other vessels.  In the 17th century, charging according to
the cargo-carrying capacity of the ship becomes dominant.
By the end of the 19th century, we find the system described by
Coase \cite{Coase2}, in which ships paid a fee (depending on
their cargo-carrying capacity) for each
entry into or exit from a harbor up to a certain number,
and nothing more for the remainder of the year.  (Such
two-part tariffs do have significant advantages in collecting
payments for services with low or zero marginal cost, as
was argued by Coase in another paper \cite{Coase1}.
We do encounter them in everyday life, for example in
restaurants that offer free refills of coffee or
other drinks.)

British lighthouse duties show growth in complexity,
but these levies did not require
knowledge of the nature of the cargo, if any.  Moreover,
government acquisition of private lighthouses,
authorized by Parliament in 1836, was stimulated by a desire
to lower the level of fees and also to make them more
uniform.  Many of the complaints were about the complexity
and the associated non-monetary costs of the old system.
(This was a time of the great movement to liberalize
trade, which reduced tariffs and other impediments.  The
famous Rhine tolls were abolished in that era as well.)

In general, lighthouse dues appear to have
been only a small part of the total cost of shipping.
Coase \cite{Coase2} cites figures for 1971-1972 that
indicate that revenues of the General Lighthouse Fund
were then on the order of two thirds of one percent
of the cost of running ships trading with the U.K.,
and thus a tiny fraction of the value of the goods
shipped.  Some rough estimates show lighthouse fees
were also low compared to cargo value or transportation
costs in earlier times.  For example, in the early 1830s,
the estimate for the total revenues of all lighthouses
(before they were nationalized and fees lowered) was
at most a quarter of a million pounds sterling (p. 47 of \cite{Prouty}),
as compared to total value of exports and imports alone (and thus
excluding the considerable domestic sea transport) of over three
hundred times as much.

It is particularly noteworthy that lighthouse charges
grew in complexity even though they were low compared
to other costs involved in shipping.  As was remarked
by Coase \cite{Coase2}, it is hard to imagine that
many decisions whether to make a voyage, or from which
harbor to make one, depended on the level of lighthouse
fees.  It would be nice to investigate this evolution
in sophistication of charging schemes, to find out what
kinds of arguments were used for and against it.  It
is likely that it was concern about fairness
(and the related issues of shipper complaints and
evasion) that dominated, as has been true historically,
and is likely to be increasingly prominent in the
future \cite{Odlyzko18,Odlyzko20,Odlyzko21}.
The pricing of goods and services of low marginal
cost is often a matter of moral philosophy.
It is not uncommon for people with no direct interest
in the subject to argue vehemently that flat rate
pricing is morally wrong because it forces light
users to subsidize heavy ones, say.

British lighthouses provide an interesting example of
pricing that was moderately complicated and intrusive.
Other countries had other policies.  The U.S., for example,
has from the beginning had lighthouse service provided by the Coast Guard,
paid for out of general federal government funds.  
A different and very interesting example is that
of the Danish Sound Tolls.  Their records have been studied
extensively, since they are unusually complete and provide
an unparalleled view of the economy of Northern Europe.
Collected in Helsingor (the Elsinore of Shakespeare's {\em Hamlet}),
the Sound Tolls evolved, primarily in the 1548-1567 period,
out of ``beaconage,'' ``a minor ship-toll in two
classes according to the sizes of the ships'' (p. 301 in \cite{Christensen})
for providing navigation beacons.  In 1548, a ``lightage''
levy of 1\% of value of cargo was imposed on goods shipped
by merchants of some countries, and in 1567 ``lastage,''
a duty based on volume measurement of ships, was introduced.
The structure and the level of the Sound Tolls kept
changing, and were very complicated, since international
treaties meant charges levied on cargo varied depending
on the nationality of the cargo owner (not that of the
ship carrying the cargo).  For more details, see the books
\cite{Christensen,Maczak}.  The interesting point is that
as the Sound Tolls' bite increased, so did their
sophistication.  Initially lastage was projected by
Danish financial leaders as a fairly simple levy,
``based on cargo capacity, with ships in ballast paying
half.''  However, customs officials persuaded the
king to modify it, so that ``from the very beginning
[it was] collected not according to number of lasts,
but to a specially elaborated {\em tariff}, in which
the rates were not only adapted according to the units of weight and measure
used, but also graduated according to the kind and value of
the merchandise'' (pp. 299-302 of \cite{Christensen}).

Modern economic concepts and models were not available
back in the 16th century, but decision makers often
did come up with ingenious schemes.  For example, 
the {\em ad valorem} levy was on the declared purchase price of the
cargo, which produced obvious incentives to falsify
documents to show low value.  To deal with such abuses,
the Danish crown had the right to  purchase goods at 
the declared value \cite{Maczak}, diminishing the incentive
to cheat.
% (and introducing what modern economists would
% call incentive compatibility to the toll system).

Compared to British lighthouse duties, the Sound Tolls
were much higher, and more intrusive, since they required
detailed information about the nature of the cargo, both
its value and its ownership.  Complaints and abuses were
frequent, and led to simplification.  First a treaty
of 1841 led to relatively uniform {\em ad valorem} 1\% toll,
and then in 1857 the Sound Tolls were abolished entirely.
The abolition was the result of several countries, 
including the U.K. and the U.S., paying a lump sum
to the Danish government.  The motivation was to free
international trade of what was perceived as a costly
encumbrance.  The Select
Committee on Sound Dues of the (British) House of Commons
in 1856 complained of the level of dues, ``but mainly [of] 
the manner in which they [were] collected.''





\section{Canals and river navigation projects}

During the Ming dynasty in China (14th to 17th centuries)
charges for commercial shipping on the Grand Canal
were a combination of simple charges based on cargo-carrying
capacity of a boat, a charge based on the value of the
cargo (initially nominally 1/30 of the market value, with
some exceptions), and a few other fees \cite{Huang}.  The {\em ad valorem}
charge was part of the general excise tax system used in
China at that time, and in many cases apparently 
``was omitted, after the tonnage was paid'' (p. 176 of \cite{Huang}).
In general, though, the excise tax was very elaborate,
and at the end of the Ming dynasty had listings for more
than 1,900 articles.  There were various irregular exactions
by officials, and ``[t]hroughout the Ming dynasty few
complaints about business taxes were directed to the
rate, most of them were about abuses and duplicated collections''
(p. 184 of \cite{Huang}).  The key point for us is that
there was extensive charging according to the value of
the cargo.
  
In England (as well as in many other countries),
canals were preceded by river navigation projects.
Government was active in two main areas.  One was in
controlling the exactions of mill and weir owners,
who often charged what were regarded as extortionate
fees \cite{Thacker,Willan}.  The other, which developed
more slowly, with
the most intensive growth starting in the early 17th century \cite{Willan},
was in authorizing navigation improvement projects.
These were usually based on permissions given to private individuals,
municipalities, or monasteries to dredge rivers, or
straighten their courses, or construct locks, in return
for the right to levy fees on boats.  There appeared to
be substantial variation in charging schemes, with a general
tendency towards schemes that are more elaborate and correspond
more with the value of service.
The very first Act of Parliament in this area was for improvements
on the river Lea in 1424.  A subsequent Act of around
1430 provided for a toll of fourpence for each ``laden
ship or boat leaving or entering the river'' (p. 469 
of \cite{Clifford}).  Later, on the
Thames (which for many centuries was a vital commercial
artery for England), tolls for using locks
were generally per ton of capacity of boat, usually for
round trip, sometimes per passage, and sometimes
if a boat returned empty, half the toll was rebated (p. 162
in \cite{Thacker}).  A lease from 1638 provided for
flat rates for  each passage up or down, but with
different rates for flat bottom boats and barges
(pp. 75-76 in \cite{Thacker}).  
The nature of the cargo, and even the volume of freight
carried, did not play a role.  With time, though, we
begin to see increased reliance on charging according
to amount of goods shipped, usually by weight.  

The evolution of tolls is nicely illustrated by the
Beverley Beck navigation project, described in \cite{Duckham}
and in much more detail in \cite{MacMahon}.
This was operated by a municipal non-profit organization,
although the revenues were sometimes used for purposes
such as street repairs, and not just for
improvement of water transport.  The first Act of Parliament,
enacted in 1727, had fairly simple tolls, although there
was variation even there, with sand charged 2 pence per ton,
and salt, iron, lead, timber, and stone 4 pence per ton
\cite{Priestley}.
(The actual list was considerably longer, and included
items such as 4 pence for each 32 firkins of butter.
All references to pence in this paper are to the old
British pence, with 240 of them to the pound.)
This was insufficient to cover the costs,
and in 1744 (according to \cite{Priestley}, and in
1745 according to \cite{Duckham}, but 
the latter source may be referring
to the date the new legislation was enacted) the navigation corporation petitioned
Parliament, asking for a more appropriate toll schedule.
The basic point they made was that ``[p]art of the trouble
lay in the tolls not being proportionate to the value
of the goods transported'' (p. 9 of \cite{Duckham}, see also
\cite{MacMahon}).
Parliament responded by granting a new charter, with
a more elaborate and generally much higher toll structure.  (It appears from
the remarks in \cite{Priestley} that a restriction was also
imposed, forbidding the use of these toll revenues
for purposes other than improvements of navigation.)  Charges for sand
remained at 2 pence per ton, those of timber, stone, and salt
were raised from 4 to 6 pence per ton, and those for iron
and lead raised from 4 to 12 pence per ton, as shown
in Table 2 \cite{Priestley}.  The change in tolls, along with
other changes in policy, and a generally
growing level of economic activity, did bring a measure of
success to the Beverley Beck project.  The interesting
point here is that this was a non-profit enterprise
that operated right on the verge of sustainability,
and a rebalancing of tolls appeared to make a
measurably positive impact.  But the case is not iron clad,
since the general increase in tolls was likely much more
important.

Canals were the next step up from river navigation projects
in complexity, cost, and efficiency.   
An interesting example is presented by the Dutch {\em trekvaarten} system
of canals and canal boats constructed for passenger transportation
in the middle of the 17th century \cite{deVries}.  It provided the
Netherlands with a communication system that for over a century
was superior to that of any other country.  Various pieces of the
system were built by agreement between pairs of cities.  Hence,
as has traditionally been common with government systems, there was
simple and inflexible pricing, with two classes of service, and
fees that did not vary much over a century and a half.
With time, though, incentives to price discriminate began to
make their mark, and provisions were made for cut-rate or even free travel 
by the poor.  Overall, though, there
was relatively little price discrimination,
and pricing was rigidly controlled by city governments.
The general rigidity of the {\em trekvaarten} system imposed
by government construction and control did lead to some missed
opportunities to make this once-novel system more competitive with
emerging alternatives, see \cite{deVries}.  

The Dutch canals were built in a very favorable terrain of a
flat country, with easy to handle soils and plenty of water.
A far more challenging project was the French Canal du Midi,
also called the Canal du Languedoc, constructed at the end
of the 17th century (during the reign of Louis XIV) to provide
an inland link within France between the Atlantic ocean and
the Mediterranean.  It was
a stunning technological achievement.  It was also a financial dud
for its government and private investors.  Still, tolls on this
canal were from the very beginning dependent on the nature
of the cargo, with an elaborate schedule of fees (pp. 193-197
of \cite{Maistre}).

Although the technology of the Canal du Midi was widely
known and admired, its poor financial results appear to
have deterred attempts to imitate it.  The modern canal
era can be said to start with the Duke of Bridgewater's
Canal in England.  Originally it was just a means of
connecting the Duke's colliery to Manchester.  The parliamentary
charter (which enabled him to take over private property,
with appropriate compensation) obliged the Duke to carry
cargo to Manchester at a maximum charge of 30 pence a ton,
and to sell his own coal in Manchester for no more than
80 pence a ton, about half the price that had prevailed
before \cite{KirkaldyE,Priestley}.  Parliament was
determined to obtain substantial benefits for the public
from the grant of government powers to the Duke.

The initial canal built by the
Duke of Bridgewater (with full credit for design
and construction usually, but incorrectly \cite{Malet}, assigned to Brindley)
was soon afterwards extended
to a canal providing a general freight connection
between Liverpool and Manchester.
This was then followed by other canals.  Interestingly
enough, in the deliberations leading to the grants
of charters for some of these projects, owners
of river navigation projects that were likely to be
adversely affected by the competition sometimes
argued that
even if a canal charter were to be granted,
it should allow only for tolls based on weight
of cargo, independent of the nature of the cargo 
(p. 101 of part 2 of \cite{Brindley} and pp. 194-200
of \cite{Phillips}):
\begin{quote}
The owners and trustees of the old navigations beg leave
to submit, ``that the tonnage, collected by the company of
proprietors, ought to be equal, through the whole canal,
for every species of goods; with some particular
exceptions.''
\end{quote}
Note that even these petitioners could not deny completely
the logic of differential pricing, and felt compelled to
allow for ``some particular
exceptions.''  At an even more basic level, charging
by the tons of cargo is already a nod in the direction
of price discrimination, since costs of operating a canal
depend only on the damage done by a boat, and that has
little relation to the cargo that is carried.
The power of price discrimination was
understood very well in those days, and was greatly feared.
There was
a frequently expressed and strong preference for simple rates.
As an example, consider the following passage
from pp. vii and viii of the Preface to \cite{Phillips}:
\begin{quote}
When the carriages which pass over a highway or a bridge,
and the lighters and barges which are used upon a navigable
canal, pay toll in proportion to their weight or their
tonnage, they pay for the maintaining [sic] those public works
in proportion to the wear and tear which they occasion.
A more equitable way of maintaining such works cannot be
found; for the tax or toll, though advanced by the carrier,
is finally paid by the consumer, to whom it is charged in
the price of his goods.
\end{quote}
But the incentives to price discriminate were powerful enough
to overcome such moral concerns.

The great financial success of the Duke of Bridgewater's
Canal 
% (which was reputed to earn profits of well over 100\% on
% invested capital in the early 19th century)
led to widespread attempts to emulate it.  In the
early 1790s, there was a canal mania, with a burst
of construction that was never to be replicated in Britain.
(The U.S. had its canal mania some decades later,
following on the great success of the Erie Canal.)
The charters of those canals show a general trend
towards greater price discrimination.  (There is
an excellent summary source on canal charters in
the compilation \cite{Priestley}.  It is hard to
do careful quantitative studies with that data,
though, until one obtains more information about
expected traffic on the various canals and the
political factors that were involved in the
parliamentary decisions.)  As described by one
historian, ``whereas the Trent \& Mersey in 1766 had
been granted tolls of [1.5 pence per ton per mile] on everything, the
Grand Junction in 1793 was given [1 pence] on merchandise,
[0.75 pence] on coal, [0.5 pence] on building material,
and [0.25 pence] on lime and limestone'' (p. 78 of
\cite{Hadfield2}).
There was also almost a universal requirement in charters
that forced canals to allow for free carriage of manure for adjacent fields
as well
as of stone for road repairs.
Unfortunately there do not appear to be any studies
of the reasons for the wide variation in toll schedules.
Were they selected with careful thought on their
economic impact, or just in response to political
pressure from local interests?  (Some of the correspondence
presented on pp. 53, 54, and 59 of  \cite{Willan2}, involving
negotiations between two towns prior to going to Parliament
for approval of a river navigation act in the 1720s suggests
that political considerations played a major role.)

Similar toll schedules depending on cargo were also
common in the United States.  As an example,
when parts of the still incomplete Erie Canal were opened
in 1820, there was a long list of tolls, concluding with
``All articles not enumerated, one cent, per ton, per mile''
(Chapter 2 of \cite{Whitford}).  The enumerated articles
(among those that were measured by the ton) were charged
tolls ranging from salt and gypsum at 0.5 cents per ton per
mile, to 1 cent for flour, to 2 cents for merchandise,
and nothing for fuel to be used in the manufacture of salt
(so that it was necessary not only to know the nature
of the cargo, but its ultimate use).

It is worth noting that canal tolls were a very substantial
part of the cost of canal transport.  There do not appear
to be any systematic studies on this subject, but there are
various snippets of information that suggest tolls often
were more than half the total cost.  (For example, p. 128
of \cite{Willan} mentions a certain canal, where transport
of a chaldron of coal was expected to cost 30 pence for
tolls and 18 pence in fees to carriers for use of the barge.)  Thus,
unlike for the telecommunications case to be discussed in
Section 8, there were high costs (to users, representing
primarily the high capital cost of construction) in the core of the
network.

Today, canal and maritime tolls still vary, and
often depend on nature of cargo.  The Erie Canal is
open again, and is used almost exclusively for recreational
purposes.  The tolls for private boats depend just on the
length of the boat.  Fee schedules for commercial vessels are
complicated, including flat annual fees.  On the Panama
Canal, tolls depend on the tonnage measurements using
that canal's own rules, with charges higher when 
carrying passengers or cargo, but not depending on the
nature of the cargo.  On the St. Lawrence Seaway, tolls
include charges depending on tonnage measurements, lock
fees, as well as levies varying with the nature of the
cargo.  For example, in 2002, Welland Canal charges
varied from Canadian \$0.6072 per ton of coal or
containerized cargo, to \$0.6956
for steel slab, and \$0.9717 for general cargo.
Curiously enough, on the Montreal/Lake Ontario section
of the Seaway, charges were lowest for coal, at \$0.541
per ton, considerably higher for containerized cargo,
at \$0.9164 per ton, and over four times as high for
general cargo, at \$2.2081 per ton.
  
Canals in England were traditionally strongly restricted to
offering their superior water transportation facilities
to carriers, and generally could not act as carriers themselves.
This prohibition was only eliminated in 1845, in
order to strengthen canal operators in their competition
with railways.  (However, as is noted in \cite{Duckham2}, for example,
there were instances of canals experimenting with
provision of direct carriage before that time.  The Duke of
Bridgewater's Canal was one of those cases.)
Thus this was an early example of the kind of structural
separation that is occasionally being proposed for
telecommunications, with a basic network operator and
other operators providing services on top of that operator's 
facilities.  However, canals did search for other
ways than those specified by their charters to increase
their profits, and to price discriminate more effectively
than allowed by their charters' toll structures.
Since actual tolls were often below the maxima specified
in the charters (as the canals maximized their revenues),
there was wide scope for varying the relative charges
on different types of goods.  (Unfortunately there do
not appear to be any studies of the extent to which
actual charges varied from the statutory ones.)
Although tolls were supposed to be uniform and proportional
to mileage, rebates offered a way around such rules.
There were many other practices by canals that were
regarded as abusive.
In particular, canals would often gain control of all
warehouses in their vicinity.  Complaints about
canal pricing, both as to the level of fees and to the extent
of discrimination, were among the major reasons for
the interest in development of railways \cite{Jackman}.
Many of the complaints were not about canal operators,
but about canal boat operators, who were generally
not regulated (except for sporadic price limitation
moves), and engaged in their own forms of price
discrimination and other widely disliked pricing
practices. 

The Duke of Bridgewater's Canal was a special case.  Its initial
phase was intended to connect the Duke's coal mines at Worsley
with the coal users in Manchester.  It thus represented the ultimate
in vertical integration, with the goods to be transported and
all transport facilities controlled by the Duke (although there
was provision for independent carriers).  This canal did provide carriage
for all goods throughout its commercial operation.  At the same
time, for many years it did encourage independent carriers,
even though there were proposals to squeeze them out.  There were
many reasons for this encouragement of the carriers, 
a major one the perception, expressed in the words of the main
manager in 1840 that ``... no company can carry as economically as an individual''
(p. 104 of \cite{Mather}).  Thus the benefits of structural separation
for the business of canal operation itself was recognized by the managers.
But the tolls on the Duke of Bridgewater's Canal did depend all along
on the nature of the cargo.  Further, as competition with railways
intensified in the late 1840s, the independent carriers were turned into
commissioned agents of the Canal (pp. 199--200 in \cite{Mather}).

While canal operators were trying to squeeze carriers
(who were trying to squeeze merchants, in ways similar
to those described below for turnpikes), carriers often
attempted to evade tolls.  They bribed toll-collectors,
misrepresented what the cargo was, or how much
there was of it, and in some cases even hid cargo
with high toll charges under commodities such as sand
for which the fees were low.  The countermeasures, just
as they are today, and would likely be in the future
with electronic communications, were based on both
technology and law.  Measurements were taken (in many
cases there were books available to canal operators,
listing canal boats, and the weight of cargo aboard
as a function of how deeply in the water they lay), and
there were punitive penalties for evasion.
For fuller discussion of this phenomenon in England
and the United States, see, for example, 
\cite{Gray,Hadfield1,Hadfield2}.

 



\section{Turnpikes and modern road transport}

British turnpikes were a controversial response to
a serious problem.  Traditionally, the King's Highway
was open to all.  The problem was how to keep it
in good condition.  As commerce grew, the need to
maintain roads became acute.  At first, in
Elizabethan times, laws were enacted compelling
all able-bodied commoner males to devote several days
a year to labor on the highways.  (See
\cite{Albert,Pawson,Webb} for references for the
background information as well as other items
below that are not attributed otherwise.)
The inequitable distribution of the burden this
imposed and the lack of effective control mechanisms
by the central government led to many complaints.
As a result, in 1663, the first turnpike was
authorized.  A local group was authorized to 
create a turnpike trust that would borrow money
to improve a section of a road, and then collect
tolls from travelers for passage over that section
of the road.  This venture was set up (as were all
subsequent turnpikes) as an ostensibly non-profit 
trust.  (There were opportunities for profits there, for
example in payment of above-market fees and other
abuses, but those were illicit, and in any case
were not the high profits that other,
more private,
enterprises, such as lighthouses and canals, offered.)
The reason for the non-profit nature of turnpikes
was presumably to allay concerns about a violation
of the ancient principle that the King's Highway
was open to all.  Still, this turnpike was very
controversial (as were many later ones). 
Apparently largely for that reason, it took until
1695 before the next turnpike was set up \cite{Albert2}.  

In the early 18th century, the turnpike movement
took off in earnest.  Although there were frequent
protests (sometimes violent, as in the burning of
the toll gates around Bristol in 1727 and 1735), by mid-1830s
there were over 20,000 miles of turnpikes in England.
As stated on p. 202 of \cite{Pawson},
\begin{quote}
The schedule of tolls chargeable by each trust was laid
down in its Act.  These varied considerably from trust
to trust, ... The schedule itself was designed to cover
all those categories of traffic which were considered
to be a charge on road repair, differentiating to a
certain extent on grounds of size and ability to pay.
Each schedule was qualified by a list of exemptions
and qualifications ...
\end{quote}
Tolls were usually doubled on Sundays for ordinary
commercial traffic, but were eliminated for travel
to or from church.  They also ``were never levied
on foot passengers, and were thus unfelt by the
labouring poor'' (p. 124 of \cite{Webb}).  There
were also options in many cases for a flat fee for
annual access.  Still,
there were countless controversies about the toll,
``the collection of which led to endless evasions,
inequalities and favouritisms of all kinds, arbitrary
exactions, and systematic petty embezzlements'' (p. 136 of \cite{Webb}).
As with canals, there do not appear to be any studies on
how the widely varying toll schedules were determined.

% In the U.S., there was a turnpike building boom
% in the first decade of the 19th century, especially
% in New England \cite{xxx}.  These enterprises were
% almost uniformly unsuccessful.

Road transport presents an interesting contrast
with canals.  Road tolls formed a much smaller
fraction of road carriers' costs than they did
for canal carriers.  It appears that in early
19th century Britain, tolls were 10-15\% of
carrier costs (Chapter 6 of \cite{Gerhold},
especially Table 14 on p. 129).  On the other
hand, feed for horses was more than half of
the costs.  Thus variable costs were very high,
which lessened incentives to price discriminate.
Also, because the wagon, the unit of transport,
was not all that expensive, shippers could
avoid the most extreme cases of differential pricing
by operating their own transport systems.  Yet
price discrimination was rife.  Chapter 7 of \cite{Gerhold}
makes for fascinating reading on this subject.  As it
notes, the most important factor was ``the principle
of charging what the traffic would bear.''  There
were variations by type of goods, and of course
by locality (reflecting competitiveness of different
markets), but ``the main distinction was between
gents' and trade price,'' and ``gents' price could be anything
from a seventh to a half above trade price'' (p. 153
in \cite{Gerhold}).  Some of this differential reflected different
costs, and often it was the result of explicit collusion.
Still, this differential existed even in competitive situations.
This provides yet another example of the phenomenon, going counter
to the prevailing doctrines in economics, law, and regulation
that price discrimination is not necessarily a sign of monopoly,
but of vigorous competition \cite{Odlyzko18}.

So far I have been producing examples of the
extensive practice of differential pricing on turnpikes
and other transport systems.  But what did it accomplish?
Perhaps it was just a way to transfer money from users
to owners, or else it might have been the result of
some mistaken dogma.  In most cases we have little
solid evidence to decide.  But there are a few examples
that demonstrate the utility of this practice.  For river 
navigation, the Beverley Beck Navigation, cited before,
showed what appears to have been a beneficial impact from
a more discriminatory toll structure.  For turnpikes, we
have some very interesting recent studies of Dan Bogart
\cite{Bogart1,Bogart2}.  They show that introduction of
turnpikes did lead to a substantial increase in spending
on roads.  (That is, the turnpike trusts, the non-profit
entities improving and maintaing the roads, and collecting
the tolls, spent considerably more on the roads than the
parishes did on their own, in response to their legal
obligation to maintain the King's Highway.)
More interestingly, the study \cite{Bogart2} demonstrates
that ``land carriage rates fell by approximately
11-16\% after turnpike trusts were adopted.''
Thus even though the free highway was replaced by a
toll road with high charges imposed on carriers, the increased
efficiency of transport on the turnpike led to end users
paying less than before.  At the same time they benefited  
from faster and more reliable service.  Thus one can safely conclude
that turnpikes were a positive contribution to society. 

Yet another interesting observation in \cite{Gerhold}
is that transport charges did not vary much with
season, even though underlying carrier costs (especially horse
feed) were extremely variable.  (For more extensive
data on this, including statistics showing the
narrowing of summer and winter fees with time, see
\cite{Bogart1,Bogart2}.  Carriers protected themselves
from the wild swings in prices by stockpiling horse
feed, but that basically meant they were absorbing the
variations.)  Thus one could say
that customers in the summer months were subsidizing
those sending goods in winter.  This tendency for
prices to vary much less than conventional economic
profit maximization would suggest has been a puzzle
to economists for a long time.  It appears to be
a response to the behavioral economics concerns
that are discussed in \cite{Odlyzko21}, and which 
limit the spread of price discrimination.

The finances of the current road transport systems are
hard to disentangle.  There are a variety of user fees
(registration, fuel taxes) as well as general tax revenues
used.  It is hard to tell just how much price discrimination
is being practiced, but it is extensive (for example, in
different charges for freight, as well as in higher profit
rates on expensive cars).  
There are two important factors to
bear in mind.  One is that most of the costs are
born by the end users directly (through purchases of cars,
paying for insurance, gas, etc.).  The other one is that
road tolls are coming back as a result of growing congestion
and improved technology.  Unlike telecommunications, where
technology is increasing capacity of fiber, coax, 
and radio transmissions, building new roads is increasingly
difficult, and making existing ones carry more traffic
can only be done to a limited extent.  At the same time,
electronic means for monitoring traffic and collecting
tolls are improving, and we see central business districts
in Norway, Singapore, and London imposing
tolls.  Most of these systems do raise privacy issues, too,
since they are centralized ones with information about
users, or at least cars.  Still, there is a strong tendency
to introduce ever more detailed monitoring of traffic,
often with the explicit goal of charging users according
to their level of activity (whether by governments or
by insurance companies).




\section{Railroads}

Railroads were the dominant industry of the 19th century.
Their similarities and contrasts with the Internet and 
the resulting implications for the economy as a whole
will be explored in \cite{Odlyzko19,Odlyzko20,Odlyzko21}.
Particularly striking are the results about privacy
and price discrimination, summarized in the
extended abstract \cite{Odlyzko18}.  In this section
I present just some brief remarks on this topic.

Early railroad charters, in both England and the U.S.,
were modeled after canal and turnpike charters,
and almost uniformly envisaged that railroad companies
would not be carriers themselves.  Instead, they were expected to
offer their facilities
for use by carriers that would carry goods and passengers in their own wagons
over the rails.  Still, these charters specified tolls
that varied greatly depending on the nature of the cargo.
(Since these were tolls for use of the rails, handling costs
were not an issue, as those were covered by carrier charges.
Any costs to the railway, such as wear and tear on the rails,
were due to the weight of the cargo,
independent of what it was.)
For example, the very first parliamentary act for a railway
was enacted in 1801.  (Previous railways had been on private
property, but in this case, as in subsequent ones, promoters
were asking for the right of eminent domain to acquire the
necessary land.)  Between the endpoints of the railway,
``chalk, lime and other manures were charged at the rate
of three-pence per ton per mile; coals, corn, potatoes, iron
and other metals, fourpence; and all goods not specified,
sixpence'' (p. 45 of \cite{Clifford}).
Thus there was no end-to-end principle, no open architecture,
and no privacy for the goods that were carried.

An interesting observation is that, just as today, the
government was trying to leave as much choice of technology
as possible
to the market (p. 54 of \cite{Clifford}):
\begin{quote}
Parliament wisely refrained from binding the first railway
projectors to adopt any specified form of rail.  Whether
a plank of wood or an iron plate should be used; whether
the rail should be laid on stone or on wooden sleepers, should
be flanged or smooth, should be flush with the ground, or 
sunk, or project above the ground, whether the wheels should
be cogged or toothed, fitting into the rail as they revolved,
to prevent skidding, or should offer a plain surface, guided
by the grooved rail:--these were questions with which
Parliament did not meddle.  Each of these plans, however,
had its advocates, and was in turn adopted.
\end{quote}
On the other hand, pricing, and especially the extent of price
discrimination, were of intense interest to Parliament,
and stringent limits were imposed in charters.
(A relevant
observation, in view of the claims that are being made today
that carriers need to have absolute control over networks, or
they won't invest in broadband, is that there did not seem to 
be any shortage of investment in canal navigation projects,
canals, turnpikes, or railroads.)
These limitations on
price discrimination, primarily through detailed toll schedules, were
pretty effective
with canals and turnpikes, but not with
railroads.

Although some railroads did operate with other companies'
equipment on their rails for decades (and modern ones do
so extensively), there was a relatively quick shift in
the 1830s and 1840s towards railroads being exclusive
carriers.  There were technical reasons promoting such
a shift (safety was jeopardized with multiple operators
and primitive technology), but there is evidence that desire
for greater control over pricing by railroads was also
a major consideration \cite{Odlyzko20}.  Once railroads
became carriers, they could engage in much more extensive
price discrimination than allowed by the toll structure
in their charters.  And, propelled by the economics of
their industry, with high fixed costs, railroads did
engage in massive price discrimination, including personal 
discrimination.  The result was massive political movements
leading to government regulation \cite{Odlyzko18,Odlyzko21}.

The full story of pricing policies of 19th century railroads
is too vast to tell here.  I will illustrate it with just
one example, that of the ``small freight wars'' of 1840-1872 in Britain.
(A much more detailed description is available in Chapter 5 of \cite{Kostal}.)
Railroads had high rates for small packages, and low
rates for large ones.  This opened up a very profitable business
for independent parcel-handlers.  Purely for purposes of illustration,
suppose that there were only two types of packages, of 50 pounds
and 1,000 pounds, and a railroad had decided that the profit-maximizing
price schedule was to 
% practice price discrimination by
charge \$1 for a 50 pound package, and \$10 for a
1,000 pound one.  An outside agent could ruin this practice
by practicing arbitrage, offering to convey 50 pound packages
for \$0.75 each, and accomplishing this by assembling them
into 1,000 pound ones, which it would then pay the railroad
\$10 to convey (or \$0.50 per 50 pound package).  That was
basically what was happening in Britain in the 19th century.
Railroads set out to fight this practice, by refusing to accept
such combined packages from independent parcel-handlers,
or by charging these shippers punitive rates.  The parcel-handlers kept
going to courts, which continued ruling in their favor and against
the railroads.  
% Quite often the courts even
% awarded the shippers payment for their legal costs.
Still, railroads
kept interfering with these carriers by making
trivial modifications to their tariffs, just enough to be
able to claim they were not doing what had been ruled
illegal in the last lawsuit they had lost.
(Those familiar with the UNE-P facilities sharing fights
over the last few years in telecommunications all
over the world may recognize
some similarities.)

The point of the example above is that railroads could not
resist the temptation to price discriminate even when it
was plainly illegal to do so.  Moreover, to do so, they had
to have knowledge of the nature of the cargo they were
carrying.

% Railroad price discrimination was extensive, and led
% to the first development of microeconomics by the
% French ``econo-engineers,'' who were trying to understand
% what the railroads were doing \cite{EkelundH}.

Eventually, railroad price discrimination led to a revolt,
and government curbs on railroad pricing practices.
The main complaint was usually less about
the level of prices (which were typically far below the
maximal levels envisaged in the charters), and more
about inequities of differential pricing \cite{Odlyzko18,Odlyzko21}.

Government regulation, imposed at the end of the 19th century,
did lead to stability for the railroad industry and prices that the
public could be persuaded were largely fair.  (It did not eliminate
differential pricing for goods, though.  The incentive for
price discrimination was too strong, and charges continued
to depend on the cargo.)  However, regulation
did strangle innovation in the industry, and resulted in gross 
inefficiencies.  With time, the arguments that the costs
of regulation were too high gained ground.
In the U.S., railroads were deregulated by the 1980 Staggers Act.
As a result, between 1984 and 2001, ``inflation-adjusted rail rates fell 45\%''
\cite{StPierre}.  However, there are still complaints about
both service quality and the degree of price discrimination.
It is estimated ``that captive shippers commonly pay rates 20\% higher 
than shippers with competitive alternatives'' \cite{StPierre}.
The federal Surface Transportation Board can intervene in 
extreme cases of gouging, but the general perception is that
competition from alternative modes of transportation is
sufficient in most cases to provide a workable transportation
system.  It is a system, though, where prices depend on
the nature of the cargo as well as local competitive conditions.




\section{Open systems and innovation}

The power to price discriminate, especially for a monopolist,
is like the power of taxation, something that can be used to
destroy.  There are many governments that are interested
in controlling Internet traffic for political or other reasons,
and are interfering (with various degrees of success) with
the end-to-end principle.  However, in most democratic societies,
the pressure to change the architecture of the Internet is
coming primarily from economic concerns, trying to extract
more revenues from users.  This does not necessarily threaten political liberty,
but it does impede innovation.  If some new
protocol or service is invented, gains from its use could
be appropriated by the carriers if they could impose special
charges for it.  

The power of price discrimination was well understood
in ancient times, even if the economic concept was not
defined.  As the many historical vignettes presented
before show, differential pricing was frequently
allowed, but only to a controlled degree.  The main
concern in the early days was about general fairness
and about service providers leveraging their control
of a key facility into control over other businesses.
Personal discrimination was particularly hated, and
preference was given to general rules applying to
broad classes (such as student or senior citizen
discounts today).  Very often bounds on charges
were imposed to limit price discrimination.

In particular, the standard economic argument for
price discrimination says that personal discrimination
(charging two otherwise identical people different
prices for the same good or service, if one values
it more highly than the other) can also provide
substantial value.  But that kind of discrimination
has traditionally been discouraged.  Discrimination
on the basis of membership in wide classes (students
or the elderly in modern societies, for example)
was often allowed, but not personal discrimination.
On canals and railroads, officially sanctioned
discrimination in freight was based on officially
prescribed categories of goods.
But sellers have often been tempted to tread in
the dangerously sensitive area of personal discrimination.

In many states in the U.S., hotels still post nominal
room rates.  This custom comes from old historical
precedents of innkeeper regulations (which were often
used as guides for canal or turnpike rules).  But
the effectiveness of such approaches is limited, since
the nominal rates can be set very high, at levels that
are practically never achieved.  There is also  no assurance
that equal rates will be charged to various hotel guests.
(Increasingly the rates that are charged vary wildly,
as hotels attempt to use airline-style ``yield-management''
techniques.)

Openness, non-discrimination, and 
% in particular
the end-to-end principle have contributed greatly to
the success of the Internet, by allowing innovation to flourish.  Service
providers have traditionally been very poor in introducing
services that mattered and even in forecasting where their
profits would come from.  Sometimes this was because of ignorance,
as in the failure of WAP and success of SMS, both of which
came as great surprises to the wireless industry, even though
this should have been the easiest thing to predict \cite{Odlyzko8}.
Sometimes it was because the industry tried to control usage
excessively.  For example, services such as Minitel have turned out
to be disappointments for their proponents largely because
of the built-in limitations.  We can also recall the
attempts by the local telephone monopolies in the mid- to
late-1990s to impose special fees on Internet access
calls.  Various studies were trotted out about the harm
that long Internet calls were causing to the network.
In retrospect, though, Internet access was a key source
of the increased revenues and profits at the local
telcos in the late 1990s.  Since the main value of the phone was its
accessibility at any time, long Internet calls led
to installation of second lines that were highly profitable for 
service providers.
(The average length of time that a phone line was in
use remained remarkably constant during that period \cite{Odlyzko2}.)

Much of the progress in telecommunications over the last
couple of decades was due to innovations by users.
The ``killer apps'' on the Internet, email, Web, browser,
search engines, and Napster, were all invented by end users,
not by carriers.  (Even email was specifically not designed
into the ARPANET, the progenitor of the Internet, and
its dominance came as a surprise \cite{Odlyzko8}.)
Today, in the wireless arena, the most vigorous growth
and innovation is coming in the license-free area,
unencumbered by high costs and inflexibility.

Yet the openness of the Internet is also perceived
as a major cause of the telecom slump.  The end-to-end
principle and the open architecture
are under assault for many reasons \cite{BlumenthalC,ClarkWSB},
but one of the major ones is that they are widely seen not to allow for
viable business models.  Email, Web, browser, search engines,
and Napster are all great, but appear not to provide any
direct revenues for carriers.
In contrast, the Minitel design was for a content
delivery network, with email
specifically not designed in.  To the extent that Minitel
succeeded,
it did so largely because of chat rooms, which were an
accidental feature designed for other purposes.  Still,
Minitel continues to be profitable for France Telecom to this
day (if one ignores the huge investment the French
government made in the development of the technology).  Hence
from the standpoint of the service providers it might
appear attractive, in spite of its limitations.
As yet another example, there are major battles over
peer-to-peer music sharing in the wireline Internet, 
with the music industry generally
fighting the ISPs (even if users are the main targets of
their assaults).  On the other hand, in the wireless area,
carriers and music companies collaborate in a mutually
profitable business of selling ring tones to users.
The architecture of cell phones allows collection of
fees from users, and content providers and carriers each get 
a negotiated slice.

Moreover, it is not just service
providers who might like to change the architecture
and the business model of the Internet.  The ``stupid
network'' is only stupid in the core, and imposes
huge burdens on end users \cite{Odlyzko3}.  Many of those
users might
be willing to sacrifice some of the openness and
flexibility in order
to be relieved of the frustrating chore of being their own
network administrators.  It is rather ironic that many
of the defenders of the open Internet castigate Microsoft
for allowing email attachments to be executed.  But that
was a feature designed to make the PC more open.

Introduction of artificial restrictions on the Internet
would be consistent with other trends in the modern
economy.  In addition to legal measures (such as DMCA),
the U.S. government is forcing major architectural
changes on the whole IT industry through the requirement for the ``digital
broadcast flag.''  The computer industry is contributing
to these trends with its development of DRM (digital
rights management) and ``trusted computing'' technologies.
The scientific and engineering developments that gave us
the openness of the PC and Internet platforms are
also enabling changes to these platforms that would
restrict what users can do with them.

It is worth noting that should 
% it turn out that 
a substantial degree of price discrimination be necessary,
some remedies that have been proposed for the problems
of the telecom industry would not be sufficient.
In particular, structural separation (splitting
the local access providers into two parts, one
that would be a monopoly provider of local connectivity,
and another one that would compete on an equal footing
with others to provide services) would likely not
be effective.  The problem is that it is the basic
connectivity piece that is expensive to provide,
whereas costs of services are lower and are decreasing
much more rapidly.  Hence it would
likely be necessary to allow the monopoly access
provider to price discriminate, and that is hard
to do effectively without control over traffic.  (Some degree
of differential pricing could be introduced without
looking at transmissions.  That is done today in the U.S.
in voice telephony,
where basic voice circuits for business use usually
cost about twice as much as for residential use.
This might not be sufficient, though.)

The general conclusion is that the open architecture
of the Internet is a major asset in promoting
innovation, but it is not guaranteed to survive
just for that reason.  However, there are other
factors that do argue in its favor.











\section{Trends in telecom pricing}

The general trend in communications has
been towards simpler pricing and decreasing price
discrimination.  This is described in detail in
\cite{Odlyzko7,Odlyzko10}.  There was one
notable counterexample, though.  The telephone
started out with flat rates almost everywhere.
The phone companies (private as well as government ones)
then fought a worldwide battle, spanning several
decades, to switch to metered billing.  They succeeded
almost everywhere, as is detailed in \cite{Odlyzko7,Odlyzko10}.
They were supported by a remarkable consensus of experts.
And indeed, the case for metering phone calls in those
days was overwhelming, since there was a high marginal
cost associated to each call, as action by an operator was
required.  The major exception to the switch to metered
rates was the United States, where, aside from a few
places such as Chicago and New York City, flat rates
were preserved for residential local calling.  This
appears to have been the result of the competition
between the Bell System and the independent phone
companies.  The need to cater to customer preferences
meant that expert opinion did not prevail, and metered
rates were not forced on everyone.

Although flat rates were regarded as damaging, they
did not harm the U.S. telephone industry in the long run, as a comparison
with other countries shows \cite{Odlyzko7,Odlyzko10}.
U.S. phone industry revenues have traditionally been higher, as
fractions of GDP, than those of most other countries,
and the industry was dynamic and profitable.  Moreover,
flat rates for local calling played a key role in the
rise of the Internet, by promoting much faster spread
of this technology in the U.S. than in other countries.
(This, as well as the FCC decisions about keeping
Internet calls free from access charges, should 
surely be added to the list of ``the 10 key choices
that were critical to the Net's success,'' that
were compiled by Scott Bradner \cite{Gillmor}.)

Today, as was already noted in the Introduction,
we are seeing the spread of flat rates to long
distance telephony and even wireless.  Yet although
telephony is evolving towards simpler pricing, it
did start out with a high degree of price discrimination
and elaborate pricing.  (That was also true of other
systems.  For example, postal
services started out with
distance-sensitive tariffs.  Later, after switching to
what are now known as ``postal rates,'' independent of
distance, they still were introducing services motivated
by the incentives to price discriminate, such as
postcards.  Moreover, although there is a version of
the end-to-end principle and open architecture
in postal systems, with first
class mail generally protected from intrusion by postal
employees, book rates involve potential inspections.
There are also stringent restrictions on competition,
and in the U.S. at least, post office boxes, although
owned by residents, can only be used for mail.)
As is detailed in \cite{Odlyzko7,Odlyzko10},
the earliest phone rates were twice as high for business
as for residential users.  (This disparity in charges
for basic monthly fees persists
in the U.S. to this day.)  Later, high long distance
rates were used to subsidize basic telephone service
and local calling.  Evening and weekend discounts
were introduced as well (although they are now
disappearing.)
There were constant attempts to limit
what customers could do, as in the Carterphone case in the 1960s
where the Bell System attempted to control what could
be interfaced with the the phone network, and even in
attempts to prevent customers from placing covers on
their telephone directories.
Hence it is possible to reconcile the view that pricing
tends to get simpler and price discrimination decreases
with advocacy of complicated pricing and extensive
price discrimination for broadband access today.  The
argument then is that such measures are necessary to develop
a new technology, even if eventually they might need to
be phased out.

While there is a trend towards simpler pricing in
telecommunications, and especially towards flat rates,
there are also attempts to increase price discrimination.
As an example, while wireline long distance voice rates are in
general dropping, most carriers' basic plan rates are increasing,
catching those who make few calls, or are ignorant or just
procrastinate.
There is also growth in other forms of first degree (personal)
price discrimination,
with wireless carriers, in particular, negotiating
with individual customers.

Interestingly enough, even though the wireless industry
is congratulating itself for avoiding the open architecture
of the Internet, it has been remarkably poor at price
discriminating.  It has done some (and in particular
has managed to charge for each handset).
But it has failed to exploit other opportunities.
It has fallen for the mirage of
mobile Internet access, and has neglected the opportunities
in providing differentiated voice services as well as
toll-free wireless calling \cite{Odlyzko8,Odlyzko11,Odlyzko17}.
Such services provide promising opportunities for drawing more
revenues from business users that are not being exploited.





\section{Telecommunications today and tomorrow}

For a discussion of the current state of the telecom
industry (with references and statistics), see
\cite{Odlyzko16,Odlyzko17}.  Here I just reiterate
a few key points, with some additional recent data.
Total telecommunications service revenues in the U.S. (and all 
statistics in
this section will be for U.S. alone) is around
\$300 billion to \$350 billion per year.  Most of that is voice,
with wireless accounting for around \$80 billion.
Although Internet backbone traffic volumes are much higher
(by a factor of more than two at the end of 2003)
than long distance voice volumes, 
Internet revenues are only about \$35 billion, with 
about \$15 billion coming from dedicated access,
and \$10 billion each from dial access and residential
broadband access.  (To be fair, it should be mentioned that
private line and Frame Relay services, which provide
intra-enterprise connectivity, bring in revenues
of about \$30 billion per year to the carriers.  
Those services are used primarily to carry IP traffic,
so they are really part of the Internet, and serve
to boost the Internet revenue figure to about \$65 billion
per year.)

Especially important is the migration of costs on
the Internet to the edges.  Of the \$35 billion
in annual Internet revenues, even the \$15 billion
for dedicated access would shrink to \$2 billion
if all the traffic coming in were aggregated into
large pipes \cite{Odlyzko16,Odlyzko17}.  
(See also \cite{Newman} for costs of broadband
access, and how little is for the basic network
infrastructure.)
The core
of the Internet, although of huge capacity, is
not expensive to run.  Technology has outrun demand,
and the entire U.S. Internet backbone traffic could
in principle be carried on a single fiber strand
\cite{Odlyzko16}.  Moreover, prices are still
dropping \cite{Jander3}.  In addition to the
illustrations in \cite{Odlyzko16,Odlyzko17}, let
us note that the transatlantic cable constructed
a few years ago by 360networks for \$850 million
was recently bought by an investor for \$18 million \cite{Berman}.
It costs about \$10 million per year to run, and 
at time of sale had lit capacity of 192 Gb/s.
(With more equipment its capacity can be
raised to several terabits per second.)  In comparison,
in early 2003, total transatlantic Internet traffic
averaged only 70 Gb/s (in the U.S. to Europe direction,
and 42 Gb/s in the reverse one \cite{Telegeography}, while
the total lit transatlantic capacity was about 2 terabits per second).
Hence, given the
distribution of demand as a function of time of day
and day of week, in principle the entire Internet 
traffic could have been
carried on that single cable, even without lighting
up additional capacity on it.

The sale price of the 360networks cable is not
an anomaly, as many other distressed carriers'
assets are being sold for pennies on the dollar \cite{Harvey}.
As a result, fiber on the main long distance routes
is essentially free.  The costs of lighting that fiber
are also relatively moderate, as the discussion of 
Cogent in \cite{Odlyzko17}
shows.  Thus the costs of long haul transport are
basically negligible compared to total telecom
spending, and are likely to remain so for the
foreseeable future.  They are so low that the
``vulture investors'' purchasing fiber assets may
not realize much value, as the major carriers may
in effect give away long distance transport to
their customers as a
way to hang onto the more lucrative services.
Comments that one hears from industry leaders,
such as that the ``competitive advantage is no longer asset
base but the sales and marketing team you've got'' \cite{Berman}
reflect this new reality. 

The fiber glut is a major contributor
to this reduction of the core of the network to
a low-cost commodity.  However, as is discussed in \cite{Odlyzko16,Odlyzko17},
it is far from being the only reason.  Technology had been
reducing costs in the core far faster
than at the edges even in the pre-Internet days.
Even if one had to build a totally new fiber network
from scratch, it would not be very expensive.  Hence,
aside from the incentive to price discriminate, the
logical and economically efficient outcome is to
run core networks as commodity providers of
a uniformly high quality service.  
Given the degree of competition on major long distance
routes and the lack of a player that might have a chance to monopolize
fiber supplies, it appears overwhelmingly likely that
core transport will continue to evolve towards a
commodity.  It will likely be profitable eventually for
one or two players, after some consolidation, but it is
unlikely to be a very large business.

On the other hand, historical precedents strongly suggest that
total telecommunications spending should resume growing again, even
when measured as a fraction of the economy \cite{Odlyzko7}.  Most of that
spending is likely to continue to be at the edges of the
network.    
It may also increasingly be in forms that do not
produce carrier revenues, as we move to customer-owned networks.
(Signs of that are the fiber strands or wavelengths that large enterprises
are increasingly purchasing to reach local exchange points,
as well as the WiFi and other home networking setups that
residential users are buying.)  There may be more heterogeneity
even in local access, with DSL, cable, and broadband
all available to most users.  

What we appear to be moving towards is
a heterogeneous mixture
of networks, unified through the Internet Protocol.
Voice will be just one of many services delivered through
a broadband link.  
The resulting system will
likely resemble what Ted Stout calls the multi-modal
transportation system, with many technologies and
specialized service providers available,
and customers selecting the best one for their needs,
often through intermediaries.  (And indeed, we are beginning
to see the emergence of such intermediaries in telecom \cite{Drogseth}.)
In such an environment, there will be a variety of players,
and there will be price discrimination, but not through
a single giant monolithic carrier, but through many
competing enterprises.  


% This might limit price discrimination
% to a tolerable degree.

% Personalized bundles





\section{Conclusions}

The historical precedents presented here, together with
basic economic arguments, help to at least partially explain
the puzzling behavior of the networking industry,
as well as of the networking research community.  They
have devoted inordinate efforts to technologies such
as ATM and QoS, even though there was abundant evidence
these were not going to succeed.  One can go further
and say that essentially all the
major networking initiatives of the last decade,
such as ATM, QoS, RSVP, multicasting, congestion pricing,
active networks, and 3G, have turned out to be duds.  
Furthermore, they all failed not because the technical
solutions that were developed were inadequate, but
because they were not what users wanted.
The misguided development of these technologies
took place because researchers and developers refused to 
take a realistic look at how networks were used, and
how they were likely to evolve.  This
behavior, though, appears to have been motivated largely
by the message they kept hearing constantly from business people
that differentiated services were a must.
The basic motive for this message 
appears to have come from the incentive
to price discriminate.  The historical evidence shows
how important a factor that has been in the past, 
especially in transportation, and
so this push is understandable.
Going forward, the incentives and the means to price discriminate
will be increasing.  This will lead to continued and 
even intensifying threats to the architecture of the Internet,
especially the end-to-end principle, and to current business
models.

Does this mean that the current architecture of the Internet
is lost?  Not necessarily.  There are several countervailing
factors.  First of all, there are general public policy 
concerns.  The end-to-end principle, and the general open
architecture of the Internet, have proven superb in stimulating
innovation and economic growth.  And even in earlier times,
the destructive powers of untrammeled price discrimination
were well understood, and usually strictly limited.
(Of course, such limits were often transgressed, most
egregiously so by 19th century railroads.  Moreover,
with the extra flexibility that modern technology provides,
there will be ever more ingenious ways to break through
barriers that governments might impose.)  Typically,
stringent limits were prescribed on tolls, dependent
on carefully chosen categories.  Thus one could make a
good case for government intervention to limit carrier practices
(as it already does with the traditional voice network).
However, based on observations of recent behavior, it
appears that governments are much more likely to listen
to appeals for closed architectures as short-term aids
to industry, than to the promise of greater payoff in the
future.
 
A more persuasive argument for government action might be that
closed architectures have
all too often been used to produce inefficiency or high profits. 
There is a growing body of evidence that privatization
and liberalization of regulations promote investment,
efficiency, and innovation (e.g., consider
\cite{DasguptaLW,FinkMR,Wallsten} and the references there).  
However, that is not necessarily a complete answer either,
since competitive markets
can be expected to produce intensified price discrimination 
\cite{Odlyzko18}, as happened with airlines and railroads
in the U.S. when regulations were liberalized in the 1980s.
Thus what would be needed is government incentives for
competition together with regulations to require open
architectures.  That might be difficult to obtain, especially
since it is precisely this combination that is seen as
having brought about the telecom crash.

Even if governments take a hands-off attitude, or promote
closed architectures, there
are other restraining influences.
One comes from the fact that content is not king \cite{Odlyzko8}.
Much of the inspiration for network designs has traditionally
come from the myth that content (in the sense of material
prepared by professionals for wide distribution, such 
as movies, or professional sports team performances) is
where the money is.  (Just consider the asymmetric bandwidth
of cable modem and DSL residential broadband links.)  
But in fact there is far more money
in providing basic connectivity.  That is what people have
always valued far more, and have been prepared to pay more
for.  (The far greater revenues of cellular carriers in
the U.S. than of cable TV providers is just one example
of this phenomenon.  For more data and arguments, see
\cite{Odlyzko8}.)  But while content delivery does lend
itself to a closed network, connectivity does not.
Open networks are likely to win because they can attract
more revenues from users.

Closely related to the false myth that content is king
is the preoccupation with real-time streaming multimedia
transmission.  That is what the networking industry has
been aiming at for decades.  Yet simple projections
(as well as evidence of actual usage) show that file
transfers are likely to dominate \cite{Odlyzko6,StArnaud}.
(This had been predicted by George Gilder 
even earlier.  Moreover, this dominance has already
occurred, even though the industry appears to be unaware
of it or its significance.  While MP3 file downloads
using various P2P services are a huge factor on practically
all networks, streaming traffic is tiny.)  But in an environment where most of the
traffic consists of file transfers, the typical methods
being planned for controlling networks do not work well.
On the other hand, such an environment does offer a
promising future for service providers.  If 
real-time streaming multimedia were to dominate, they
would face a bleak future.  As soon as residential
connection speeds reached a few tens of megabits
per second, there would be nothing new for them
to offer in terms of basic connectivity.  
However, when file transfers dominate,
and faster than real-time transfers of videos is
what matters to users, service providers can
keep endlessly upgrading their customers' connections,
and use increasing speeds as a market segmentation
device.  The significance of the low utilization
of data networks \cite{Odlyzko4,Odlyzko6} is that
what matters to users is not getting lots of bits, but
% (and indeed, by the official FCC definition
% of broadband, we all have broadband courtesy
% of the postal service \cite{Odlyzko17}), but
getting a moderate amount of bits quickly, in
other words low transaction latency. 
What this says, in effect, is that the networking
industry is trying to do differential pricing
in the wrong way, along the wrong dimension.
Eventually they will learn the error of their
ways, but it might be a long time until they do.

% There are additional ways for the networking industry
% to segment the market.  Given the migration of costs
% to the edges of the network \cite{Odlyzko3}, there
% are ways to provide services to customers on their
% premises, by outsourcing their network operations.

There are some other factors that argue for a
comparatively open network.
The increasing heterogeneity of
the telecommunications network (as well as its increasing
importance) will mean that users will be able to mitigate
restrictive practices by bypassing service providers, as
large enterprises are beginning to do by buying their
own fiber.  This bypass strategy will be facilitated by the
general migration of costs and complexity to the edges.  
The rapid advances in fixed wireless threaten to make
fiber-to-the-home unnecessary \cite{Odlyzko17}, and also
to destabilize what might have developed into a cozy duopoly
of DSL and cable modem providers.  It will not be necessary
for fixed wireless to grab the lion's share of the Internet
access market.  It will suffice if it has a small share, but
is available as a viable
alternative to DSL and cable modems.  (The much better
scaling properties of wireless, without the huge initial
costs that are proportional to the number of potential
customers, as opposed to actual ones, are what makes this
disciplining role possible.)
We are likely
to end up with a system like the multi-modal transportation
system of today, which is rife with discriminatory practices
(just think of the variation in prices by household moving
companies), but where such practices are limited to a
tolerable degree.

A key distinction between the Internet and the transportation
systems discussed earlier is that in transport, most of the
costs were associated with the core of the network.  On the
Internet, on the other hand, the complexity, costs, and therefore revenue
opportunities are largely at the edges.  It is hard in such
a situation to design a network architecture that will 
provide necessary controls for carriers.
Furthermore (and this is very important in view of the
discussion later about behavioral economics constraints),
transportation charges tended to be levied relatively
infrequently, on the carriers, and therefore users did not
have to interact with them directly.

Yet another distinction is that in transportation,
the nature and value of goods does not change very rapidly.
In the new economy with extensive information goods,
though, that is not the case.

Perhaps the most potent limitation on the proposed new
architectures for the Internet, and the associated
discriminatory practices, is posed by a range of
factors deriving ultimately from behavioral economics.
People react extremely negatively to price discrimination.
They also dislike the bother of fine-grained pricing,
and are willing to pay extra for simple prices, especially
flat-rate ones.  Furthermore (and this is specific to
telecommunications and other information goods industries,
and does not apply to transportation), technology is
rapidly increasing available bandwidth, so the primary
imperative for service providers is to persuade their
customers to increase their usage.  (The suggestion
in \cite{Odlyzko17} that telecom service providers
buy out music studios and offer recorded music for
free with their broadband connections was only
slightly tongue-in-cheek.  The reviled peer-to-peer
traffic is a major stimulant of the demand for
broadband.)
Constraining
architectures and pricing structures work against
increased usage.  For more details, see \cite{FishburnOS,Odlyzko7,Odlyzko10},
and (for a short version with pointers to the literature)
\cite{Odlyzko15a}.  The general conclusion of
\cite{Odlyzko7} was that price discrimination and
fine-grained pricing are likely to prevail for goods
and services that are expensive and bought infrequently.
For purchases that are inexpensive and made often,
simple pricing is likely to prevail.  And we see
these trends in the history of transportation.
For example, as is detailed in \cite{Odlyzko21},
in the first half of the 19th century, railroads in
England were limited in attempts to practice explicit
price discrimination by law and lack of proper technology
(such as the positive passenger identification that
airlines rely on today).  Hence they resorted to extreme
forms of versioning, running third class passenger cars
without roofs, and sometimes even without seats.  
These practices led to the famous 1844 law that
forced the railroads to run the so-called ``Parliamentary
trains,'' something they did only under duress and with
loud protests.  This was in the early days of the
industry, when railroad travel was rare and expensive.
By the end of the 19th century, the overwhelming majority
of passengers traveled in 3rd class cars, which by that
time were much more comfortable.
(And even then, just as today, operators had great
difficulty figuring out how their business was evolving.
As was noted by an observer in 1885, at that time
``[r]ailway managers had not yet discovered that
third class traffic was their main stay,'' p. 100 of
\cite{Clifford}.)
Now the Internet 
already pervades society, and will be even more ubiquitous
in the future, used round the clock in a variety of
applications.  Simplicity is likely to be key to
acceptance.  We see this phenomenon in operation today.
Although hotels, golf courses, and other service
providers are rushing to imitate airline yield
management systems, many are discovering that this
is not necessarily the way to riches. For example,
Amtrak, after extensive experimentation with
complicated pricing, has decided to pull back to
the traditional approaches of simple and stable
fares \cite{Machalaba}.  Hence telecom service
providers are likely to discover that the elaborate
architectures they are dreaming of will work against
their interests.

The general conclusion then is that the historical
record of the transportation industry does demonstrate
the importance and prevalence of discriminatory policies
that are incompatible with the basic architecture of
the current Internet.  This probably accounts for much of the
push to build new networks, or modify the current ones
so as to provide more control for service providers
over what customers do.  However, the Internet is special,
in its importance as an enabler for the rest of the
economy, in its migration of costs and capabilities
to the edges, in its primary value being in connectivity
and low transaction latency, and in its pervasiveness and frequency
of use.  Hence in spite of the strong push from the
industry, there are good prospects that the open
architecture of the Internet will survive.











\section*{Acknowledgments}

Sasha Nichols-Geerdes provided extensive and valuable
research assistance in locating historical sources.
Jouko Tossavainen helped with pointers to information 
about the Sound Tolls.  Ted Stout provided helpful
perspectives on the transportation industry, and
Sam Paltridge references to telecom liberalization
analyses.  
Ross Anderson,
Dan Bogart,
Bob Briscoe,
Liudvikas Bukys,
Dave Burstein,
Steve Crandall,
Daniel Davis,
Rolf Engstrand,
Bob Frankston,
Jim Gray,
Tom Hazlett,
Chris Hogendorn,
Tim Janik,
Ihor Lys,
Paul Odlyzko,
Andy Oram,
Hal Purdy,
Jere Retzer,
Nate Taylor,
Adam Thierer,
and 
Philip Webre
helped with comments on an
earlier draft.



\begin{thebibliography}{4}
%
\bibitem{Albert}
W. Albert, {\em The Turnpike Road System in England, 1663-1840,}
Cambridge Univ. Press. 1972.
%
\bibitem{Albert2}
W. Albert, ``Popular opposition to turnpike trusts in early eighteenth
century England,'' {\em J. Transport History,} ser. 2, vol. 5 (1979),
pp. 1-17.  See also ``Debates in transport history: Popular opposition
to turnpike trusts?'' by E. Pawson and W. Albert in
{\em J. Transport History,} ser. 3, vol. 5 (1984), pp. 57--68.
%
\bibitem{AnaniaS}
L. Anania and R. J. Solomon, Flat--the minimalist price, pp.  91-118 in 
{\em Internet Economics,} L. W. McKnight and J. P. Bailey, eds., MIT Press, 1997.  
Preliminary version in {\em J. Electronic Publishing,} special issue on 
Internet economics, $\langle$http://www.press.umich.edu/jep/$\rangle$.
%
\bibitem{Berman}
D. K. Berman, ``Telecom investors envision potential in failed networks,''
{\em Wall Street J.,} Aug. 14, 2003.
%
\bibitem{Bhagavath}
V. K. Bhagavath, with C. Rutstein, H. Liddell, and A. Tseng,
``How to make VoIP successful,'' Forrester Research Report,
Dec. 2002.
%
\bibitem{BlumenthalC}
M. S. Blumenthal and D. D. Clark, ``Rethinking the design of the Internet:
The end to end argument vs. the brave new world,'' {\em ACM Transactions
on Internet Technology,} {\bf 1}, no. 1, Aug. 2001, pp. 70-109.  Available
at $\langle$http://cfp.mit.edu/materials.html$\rangle$.
%
\bibitem{Bogart1}
D. Bogart, ``Institutional innovation and infrastructure investment: An
evaluation of the turnpike system in eighteenth century England,''
Nov. 26, 2003 preprint, available at
$\langle$http://aris.ss.uci.edu/econ/paper/2003-04/Bogart-02.pdf$\rangle$.
%
\bibitem{Bogart2}
D. Bogart, ``Turnpike trusts and the transportation revolution in
eighteenth century England,'' Jan. 7, 2004 preprint, available at
$\langle$http://orion.oac.uci.edu/$\sim$dbogart/transport\_revolution\_feb12\_two.pdf$\rangle$.
%
\bibitem{Bradner}
S. Bradner, ``Will there be a next-generation network?,''
{\em Network World,} July 21, 2003.  Available at
$\langle$http://www.nwfusion.com/columnists/2003/0721bradner.html$\rangle$.
%
\bibitem{Brindley}
J. Brindley, {\em The History of Inland Navigations, Particularly Those
of the Duke of Bridgewater, in Lancashire and Cheshire, and the Intended
one Promoted by Earl Gower and Other Persons of Distinction in
Staffordshire, ...,} London, Printed for T. Lowndes, 1766.
%
\bibitem{ClarkWSB}
D. D. Clark, J. Wroclawski, K. R. Sollins, and R. Braden,
``Tussle in cyberspace: Defining tomorrow's Internet,"  {\em ACM SIGCOMM 2002,}
available at $\langle$http://www.acm.org/sigcomm/sigcomm2002/papers/tussle.html$\rangle$.
%
\bibitem{Christensen}
A. E. Christensen, {\em Dutch Trade to the Baltic About 1600: Studies
in the Sound Toll Register and Dutch Shipping Records,} E. Munksgaard,
Copenhagen, and Martinus Nijhoff, The Hague, 1941.
%
\bibitem{Clifford}
F. Clifford,, {\em A History of Private Bill Legislation,} vol. 1,
Butterworths, 1885.
%
\bibitem{Coase1}
R. H. Coase, ``The marginal cost controversy,'' {\em Economica,} N.S. {\bf 13},
no. 51, Aug. 1946, pp. 169--182.  
%
\bibitem{Coase2}
R. H. Coase, ``The lighthouse in economics,'' {\em J. Law and Economics,} 
{\bf 17}, no. 2, Oct. 1974, pp. 357-376.  Reprinted in R. H. Coase,
{\em The Firm, the Market, and the Law,} Univ. Chicago Press, 1988,
pp. 187-213.
%
\bibitem{CoffmanO1}
K. G. Coffman and A. M. Odlyzko, ``The size and growth rate of the
Internet,''
{\em First Monday,} {\bf 3}, no. 10, Oct. 1998,
$\langle$http://firstmonday.org/$\rangle$.
Also available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko$\rangle$.
%
\bibitem{DasguptaLW}
S. Dasgupta, S. Lall, and D. Wheeler, ``Policy reform, economic growth, and
the digital divide: An econometric analysis,'' 
Aavailable at
$\langle$http://econ.worldbank.org/files/1615\_wps2567.pdf$\rangle$.
%
\bibitem{Davis}
D. Davis, ``Shine your light on me ...,'' Dec. 23, 2002 blog entry,
$\langle$http://d-squareddigest.blogspot.com/2002\_12\_22\_d-squareddigest\_archive.html\#86435321$\rangle$.
%
\bibitem{deVries}
J. de Vries, {\em Barges and Capitalism: Passenger Transportation in the Dutch Economy, 1632-1839},
HES Publishers, 1981. 
%
\bibitem{Drogseth}
D. Drogseth, ``New providers manage bandwidth acquisition,''
{\em Network World,} Oct. 6, 2003.  Available at
$\langle$http://www.nwfusion.com/newsletters/nsm/2003/1006nsm1.html$\rangle$.
%
\bibitem{Duckham}
B. F. Duckham, {\em The Inland Waterways of East Yorkshire, 1700-1900,}
East Yorkshire Historical Society, 1972.
%
\bibitem{Duckham2}
B. F. Duckham, ``Canals and river navigation,'' pp. 100--141 in
{\em Transport in the Industrial Revolution,} D. H. Aldcroft and
M. J. Freeman, Manchester Univ. Press, 1983.
%
\bibitem{EkelundH}
R. B. Ekelund, Jr., and R. F. Hebert, {\em Secret Origins of Modern
Microeconomics: Dupuit and the Engineers,} Univ. Chicago Press, 1999.
%
\bibitem{Epstein}
R. A. Epstein, ``The libertarian quartet,'' {\em Reason}, Jan. 1999, available at
$\langle$http://reason.com/9901/bk.re.thelibertarian.shtml$\rangle$.
%
\bibitem{FinkMR}
C. Fink, A. Mattoo, and R. Rathindran, ``An assessment of telecommunications reform
in developing countries,'' 
World Bank Policy Research Working
Papers, no. 2909, October 11, 2002, available at
$\langle$http://econ.worldbank.org/resource.php?type=5\&id=20745$\rangle$.
%
\bibitem{FishburnOS}
P. C. Fishburn, A. M. Odlyzko, and R. C. Siders,
``Fixed fee versus unit pricing for information goods: competition, 
equilibria, and price wars,'' {\em First Monday,} vol. 2, no. 7, July 1997,
$\langle$http://firstmonday.org/issues/issue2\_7/odlyzko/index.html$\rangle$.
Revised version in {\em Internet Publishing and Beyond: The Economics of
Digital Information and Intellectual Property,} B. Kahin and H. R. Varian, eds., 
MIT Press, 2000, pp. 167-189.  Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko$\rangle$.
%
\bibitem{Gerhold}
D. Gerhold, {\em Road Transport Before the Railways: Russell's London
Flying Waggons,} Cambridge Univ. Press, 1993.
%
\bibitem{Gillmor}
D. Gillmor, ``10 choices that were critical to the Net's success,''
{\em The Mercury News,} Sept. 8, 2002.  Available at
$\langle$http://www.siliconvalley.com/mld/siliconvalley/business/columnists/4029770.htm$\rangle$.
%
\bibitem{Gray}
R. D. Gray, {\em The National Waterway: A History of the Chesapeake and
Delaware Canal, 1769-1985,} Univ. Illinois Press, 1989.
%
\bibitem{Grierson}
J. Grierson, {\em Railway Rates: English and Foreign,} Edward Stanford, 1886.
%
\bibitem{Hadfield1}
C. Hadfield, {\em British Canals: An Illustrated History,} 2nd ed., 3rd printing,
David \& Charles, 1966.
%
\bibitem{Hadfield2}
C. Hadfield, {\em The Canal Age,} David \& Charles, 1968.
%
\bibitem{Harvey}
P. Harvey, ``Tyco to unload telecom assets,'' {\em  Lightreading,}
Nov. 5, 2003.  Available at
$\langle$http://www.lightreading.com/document.asp?doc\_id=43047$\rangle$.
%
\bibitem{Huang}
R. Huang, {\em The Grand Canal During the Ming Dynasty, 1368-1644,} Ph.D. Thesis,
Univ. Michigan, 1964.
%
\bibitem{Isenberg}
D. Isenberg, ``Rise of the stupid network: Why the Intelligent
Network was once a good idea, but isn't anymore.  One telephone company
nerd's odd perspective on the changing value proposition,'' {\em Computer
Telephony,} August 1997, pp. 16, 18, 20, 24, 26.  Available at
$\langle$http://www.manymedia.com/david/stupid.html$\rangle$.
Revised version, ``The dawn of the stupid network,'' in {\em ACM netWorker,}
vol. 2, no. 1 (Feb.-Mar. 1998), pp. 24-31.
%
\bibitem{Jackman}
W. T. Jackman, {\em The Development of Transportation in Modern England,}
Cambridge Univ. Press, 1916.
%
\bibitem{Jander3}
M. Jander, ``Bandwidth prices sinking, says report,'' {\em  Lightreading,} 
October 6, 2003.  Available at
$\langle$http://www.lightreading.com/document.asp?doc\_id=41299$\rangle$.
%
\bibitem{KirkaldyE}
A. W. Kirkaldy and A. D. Evans, {\em The History and Economics of
Transport,} Pitman \& Sons, 1915.
%
\bibitem{Kostal}
R. W. Kostal, {\em Law and English Railway Capitalism, 1825-1875,}
Oxford Univ. Press, 1994.
%
\bibitem{Machalaba}
D. Machalaba, ``Amtrak timetables to feature ticket prices, simpler format,''
{\em Wall Street J.,} Aug. 6, 2003.
%
\bibitem{MacMahon}
K. A. MacMahon, ``Beverley and its Beck: Borough finance and a town 
navigation 1700-1835,'' {\em Transport History,} vol. 4, no. 2, July 1971, pp. 121--143.
%
\bibitem{Maczak}
A. Maczak, {\em Miedzy Gdanskiem a Sundem: Studia nad handlem baltyckim
od polowy XVI to polowy XVII wieku,} Pan. Wyd. Naukowe, Warsaw, 1972.
%
\bibitem{Maistre}
A. Maistre, {\em Le Canal des Deux Mers: Canal Royal du Languedoc 1666-1810,}
Edouard Privat, Toulouse, 1968.
%
\bibitem{Maitland}
J. Maitland, ``Carriers resist VOIP peering,'' {\em Light Reading,}
Dec. 17, 2003.  Available at
$\langle$http://www.lightreading.com/document.asp?doc\_id=45088$\rangle$.
%
\bibitem{Malet}
H. Malet, {\em Bridgewater: The Canal Duke, 1736-1803}, Manchester
Univ. Press, 1977.
%
\bibitem{Marvin}
C. Marvin, {\em When Old Technologies Were New: Thinking About Electric
Communication in the Late Nineteenth Century}, Oxford Univ. Press. 1990.
%
\bibitem{Mather}
F. C. Mather, {\em After the Canal Duke: A Study of the Industrial Estates
Administered by the Trustees of the Third Duke of Bridgewater in the Age
of Railway Building 1825-1872,} Oxford Univ. Press, 1970.
%
\bibitem{Newman}
S. Newman, ``Broadband access platforms for the mass market: An
assessment,'' presented at 2003 Telecommunications Policy Research
Conference, available at
$\langle$http://intel.si.umich.edu/tprc/papers/2003/254/BbandAccessPlatforms.pdf$\rangle$.
%
\bibitem{Odlyzko2}
A. M. Odlyzko, ``The economics of the Internet: Utility, utilization,
pricing, and Quality of Service,'' 1998 unpublished manuscript, available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko3}
A. M. Odlyzko, ``Smart and stupid networks: Why the Internet is like
Microsoft,'' {\em ACM netWorker} {\bf 2}, no. 5, Dec. 1998, pp. 38-46.  Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko4}
A. M. Odlyzko, ``Data networks are mostly empty and for good reason,''
{\em IT Professional} {\bf 1}, no. 2, March/April 1999, pp. 67-69.  Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko5}
A. M. Odlyzko, ``The current state and likely evolution of the Internet,''
in {\em  Proc. Globecom'99,} pp. 1869-1875, IEEE, 1999.  Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko6}
A. M. Odlyzko, ``The Internet and other networks: Utilization rates
and their implications,'' {\em Information Economics \& Policy}, {\bf 12}
(2000), pp. 341-365.
(Presented at the 1998 Telecommunications Policy Research Conference.)
Also available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko7}
A. M. Odlyzko, ``The history of communications and its implications for the
Internet,'' 2000 unpublished manuscript,
available at 
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko8}
A. M. Odlyzko, ``Content is not king,''
{\em First Monday,} {\bf 6}, no. 2, February 2001,
$\langle$http://firstmonday.org/issues/issue6\_2/odlyzko/$\rangle$.
Also available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko10}
A.M. Odlyzko, ``Internet pricing and the history of communications,''
{\em Computer Networks,} {\bf 36} (2001), pp. 493-517.  Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko11}
A. M. Odlyzko, ``Talk, Talk, Talk: So who needs streaming video
on a phone? The killer app for 3G may turn out to be--surprise--voice
calls,'' {\em Forbes,} August 20, 2001, p. 28.  Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko14}
A. M. Odlyzko, ``Internet TV: Implications for the long distance
network,'' in {\em Internet Television,}
E. Noam, J. Groebel, and D. Gerbarg, eds., Lawrence Erlbaum Associates, 2003, pp. 9-18.
(Proceedings of workshop held at Columbia University in Nov. 2000.)
Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko15a}
A. M. Odlyzko, ``The case against micropayments,'' in {\em Financial 
Cryptography: 7th International Conference, FC 2003,} R. N. Wright, ed., 
Lecture Notes in Computer Science \#2742, Springer, 2003, pp. 77-83.
Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko16}
A. M. Odlyzko, ``Internet traffic growth: Sources and implications,''
{\em Optical Transmission Systems and Equipment for WDM Networking II,}
B. B. Dingel, W. Weiershausen, A. K. Dutta, and K.-I. Sato, eds., {\em Proc. SPIE,}
vol. 5247, 2003, pp. 1-15.  Available at
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko17}
A. M. Odlyzko, ``The many paradoxes of broadband,'' {\em First Monday,}
vol. 8, no. 9, September 2003,  
$\langle$http://firstmonday.org/issues/issue8\_9/odlyzko/index.html$\rangle$.
Also available at 
$\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko18}
A. M. Odlyzko, ``Privacy, economics, and price discrimination on the Internet,''
{\em ICEC2003:  Fifth International Conference on
Electronic Commerce,} N. Sadeh, ed., ACM Press, 2003, pp. 355-366.
Available at $\langle$http://www.dtc.umn.edu/$\sim$odlyzko/doc/recent.html$\rangle$.
%
\bibitem{Odlyzko19}
A. M. Odlyzko, ``The Internet and 19th century railways,'' manuscript in preparation.
%
\bibitem{Odlyzko20}
A. M. Odlyzko, ``Privacy, price discrimination, and the future of ecommerce,'' manuscript in preparation.
%
\bibitem{Odlyzko21}
A. M. Odlyzko, ``Fairness and price discrimination on the {I}nternet 
and on 19th century railroads,'' manuscript in preparation. 
%
\bibitem{Pawson}
E. Pawson, {\em Transport and Economy: The Turnpike Roads of Eighteenth
Century Britain,} Academic Press, 1977.
%
\bibitem{Phillips}
J. Phillips, {\em A General History of Inland Navigation, Foreign and
Domestic; Containing a Complete Account of the Canals Already Executed
in England, with Considerations ...,} London, Printed for I. and J. Taylor, 1792.
%
\bibitem{Priestley}
J. Priestley, {\em  Historical Account of the Navigable Rivers, Canals and 
Railways Throughout Great Britain,} 2nd ed., Cass, 1967.  (Original edition 1831.)
%
\bibitem{Prouty}
R. Prouty, {\em The Transformation of the Board of Trade, 1830--1855: A Study
of Administrative Reorganization in the Heyday of Laissez Faire,} William Heinemann,
1957.
%
\bibitem{StPierre}
N. St. Pierre, ``Railroads: Asleep at the switch:  Lousy
service is driving away freight customers,'' {\em Business Week,}
April 2, 2001.
%
\bibitem{SalzerRD}
J. Salzer, D. Reed, and D. D. Clark, ''End-to-end arguments in system
design,'' {\em ACM Transations on Computer Systems,} {\bf 2}, no. 4,
Nov. 1984, pp. 277-288.  Available online at
$\langle$http://web.mit.edu/Saltzer/www/publications/endtoend/endtoend.txt$\rangle$.
%
\bibitem{Standage}
T. Standage, ``World telecoms: Beyond the bubble,'' {\em The Economist,}
Oct. 16, 2003.  Available online at
$\langle$http://www.economist.com/displaystory.cfm?story\_id=2098913$\rangle$.
%
\bibitem{StArnaud}
B. St. Arnaud, ``The future of the Internet is NOT multimedia,'' {\em Network
World}, Nov. 1997.  Available at
$\langle$http://www.canarie.ca/$\sim$bstarn/publications.html$\rangle$.
%
\bibitem{Telegeography}
``Global Internet Geography 2004,'' TeleGeography, 2003.
%
\bibitem{Thacker}
F. S. Thacker, {\em The Thames Highway: Vol. I: General History,}
1st ed. 1914, reprinted with a new introduction by C. Hadfield,
David \& Charles, 1968.
%
\bibitem{Trethewey}
K. Trethewey, {\em The Background to Lighthouse History in England, 
Ireland, Scotland and Wales,} available at
$\langle$http://www.btinternet.com/$\sim$k.trethewey/genindex.html$\rangle$.
%
\bibitem{Varian}
H. R. Varian, ``Markets for public goods?,'' {\em Critical Reviews,}
{\bf 4}, no. 4, 1994, pp. 539-557.  Available at
$\langle$http://www.sims.berkeley.edu/$\sim$hal/public.pdf$\rangle$.
%
\bibitem{Wallsten}
S. J. Wallsten, ``Telecom traffic and investment in developing countries: The effects
of international settlement rate reductions,'' World Bank Policy Research Working
Papers, no. 2401, July 2000, available at
$\langle$http://econ.worldbank.org/resource.php?type=5\&id=2381$\rangle$.
%
\bibitem{Warsh}
D. Warsh, economicprincipals.com column, Sept. 29, 2002.  Available at
$\langle$http://www.economicprincipals.com/issues/02.09.29.html$\rangle$.
%
\bibitem{Webb}
S. Webb and B. Webb, {\em English Local Government: The Story of the
King's Highway,} Longmans, 1913.
%
\bibitem{Whitford}
N. E. Whitford, {\em History of the Canal System of the State of New York
Together with Brief Histo