The Internet: A Paradigmatic Rupture in Cumulative Telecom Evolution.

The Internet: a Paradigmatic Rupture in
Cumulative Telecom Evolution
PETROS KAVASSALIS, RICHARD JAY SOLOMON AND
PIERRE-JEAN BENGHOZI1
. (Massachusetts Institute of Technology, Research Program on
Communications Policy (RPCP), Cambridge, MA, USA and 1CNRS, Centre
de Recherche en Gestion (CRG) de l'Ecole Polytechnique, Paris, France)
The paper investigates the impact the Internet may have in the evolution of
telecommunications networks. First, we show why the Internet, emerging from a
different cognitive perception of the data communication problem, has led to a new
network architecture based on: (i) the distribution of the 'network intelligence' to the
user equipment; (ii) the very cost-effective 'statistical sharing' of the network resources
(i.e. getting the whole bandwidth of the network fbr short periods of time); (Hi) the
establishment of an Internet Protocol (IP) 'gateway' facilitating interoperability
between heterogeneous infrastructure facilities—instead of the operator-controlled
homogeneity of the telecom networks; and (iv) an 'adaptative' way for open
standards-setting. Second, we suggest that two technological trajectories (telecom—
'creative accumulation' and Internet—'creative destruction') should dynamically
co-exist henceforth and compete for market shares—possibly during later evolutionary
stages generating relatively different national or even localized (e.g. local providers)
trajectories of evolution (with differing interfaces and standards). Furthermore, we
2 explore the question of whether the Internet's interoperability model may be a useful
•* policy paradigm for future information infrastructures, and we start to discuss the
I implications of requisite interoperability on the communications industry's structure
z
itself. Overall, our preliminary observations raise questions about the possibilities of two
"2 'technological trajectories' co-existing, and the relationship between the interoperability
J and learning conditions in the network industries.
© Oxford University Press 1996
1097

The Internet: a Paradigmatic Rupture in Cumulative Telecom Evolution
1. Introduction
The 'Problem'
It is commonly held that the standards-setting process in the
telecommunications industry suffers from enormous inefficiencies and delays
that make it difficult to keep up with a rapidly changing industry. Traditional
telecom standards-making bodies magnify the Tempus syndrome (Simon,
1969). Tempus, a hard-working manufacturer of exquisite watches, was
jealous at the prospering business of his neighbour Hora. Nevertheless, both
Tempus and Hora enjoyed a good reputation and new orders were constantly
arriving, usually by telephone, at both shops. But whereas Hora responded
successfully to customers demands, Tempus' production line failed, disturbed
by the frequent telephone calls for new orders. The two high-quality
technicians had radically different approaches to assembling their 1000-piece
watches. Tempus put his watches together piece-by-piece. Hora, whose
watches were less complex than those of Tempus, made ten-piece
sub-components that could be assembled into a completed watch as orders
came in. Telephone orders interrupting Hora's work made him lose only a
small part of his effort, but interruptions were disastrous for Tempus, whose
entire watch assembly would collapse even if he had 995 parts already in
place. As Eldredge (1995), who recalls Simon's story, concludes, 'the more
popular {Tempus'] work, the more the phone would ring, and the more
difficult it became for him to finish a single watch...',
'Old' and 'New' Models for Standard-setting
The Tempus-Hora metaphor highlights the differences between the old
standardization model, as followed by most standards bodies today, and the
new approach exemplified by the Internet.1 Traditional standards processes
and products, even if they try to adapt to the digital multilayer paradigm
requirements,2 have not done well in the marketplace in spite of considerable
investments.
' For a more detailed analysis of the differences between 'old' and 'new' models of IT standards-making,
see: Solomon and Rutkowski (1992). Furthermore, Kahin and Abbate (1995) propose an extensive analysis
of the conflicting tendencies surrounding current standard-setting processes in the communication
networks industries.
1
Digital networks are represented as a multi-level system, where the different levels (layers in the
network parlance) accomplish different performances and where protocols help to organize the interaction
between them. There are two major protocol suites, OS1 and TCP/IF) produced by radically different
processes. Generally speaking, the telecommunications industry has mostly adopted the OSI 'cognitive'
— 1098

The Internet: a Paradigmatic Rupture in Cumulative Telecom Evolution
Let us give some examples. The huge technological battles of the 1980s are
ending but are leaving an impression of'much ado about nothing". Integrated
Services Digital Network (ISDN)—considered as the next step of the
digitalization of the telephone network—has been only a moderate success
where private lines are forbidden or prohibitively expensive. Where the
private data networking has always been a significant part of the market, as
in the USA, ISDN is just a transitional technology for access to private
systems such as the Internet access providers' networks. Similarly, the
Intelligent Network (IN) concept3 for conventional central offices and the
X25 (the network protocol used by public telecommunications operators for
providing data services) packet protocol have been superceded in functionality
and rapid diffusion by the Internet's TCP/IP protocols.4 Either the standards
turn out to be sub-optimal despite the enormous designs effort in advance of
application, as with IN, or too late, as in ISDN. Or, as in the case with X25,
the approach itself—top-down, too close to the telecom carriers' 'operational
perceptions':—has become increasingly unworkable for advanced, distributed
processing networks. In all of these examples, the standards makers failed to
anticipate the single, most influential change: the advent of incredibly cheap
and very powerful microprocessors embedded in Customer Premise
Equipment (CPE).
To summarize, the problem stems from the cognitive limits of the 'old
model' standards-setting organizations: 'the old model is the traditional
hardware oriented approach followed by standards organizations...where
signal control passes directly from one end of a closed system to the other...'
(Solomon and Rutkowski, 1992). The old model worked well when trans-
mission characteristics dominated interface specifications. As transmission
moved from analogue to digital, from static to dynamic through computer
processing, and as switching became more distributed, the old model gave
only the illusion of being sufficient. Or, in order to be sufficient, the functional
levels of the multilayer paradigm had to be squeezed into a centralized
implementation whereby: (i) all intermediary switches along a routing path
are responsible for strongly controlling applications semantics; and (ii)
customer's network configuration is strongly dependent on an out-of-band
framework while the Internet it the outcome of the widespread implementation of the TCP/IP
protocols (see below).
' The label 'intelligent network' signifies the ongoing upgrading of the PSTN (Public Switched
Telephone Network) envisaged to add more switching functionality (computerized databases and very
complex software).
4
As we have already mentioned, TCP (Transmission Control Protocol) and IP (Internet Protocol) are the
main protocols supporting the Internet. As we will demonstrate later in this paper, X25 and TCP/IP are
building on competing networks concepts.
1099

The Internet: a Paradigmatic Rupture in Cumulative Telecom Evolution
signalling system, residing in extremely sophisticated and expensive
databases—more or less controlled by the network operator. This has been
increasingly irrelevant to the emerging networking world where applications
want to be functionally independent from infrastructure facilities and
technologies.
Indeed, the Internet indicates that the technological potential of the digital
paradigm is larger than within the traditional telecom approach. The Internet
is the archetype of a model where (i) cognizance of applications semantics is
strictly reserved to the users' terminals; and (ii) an intermediary layer, the IP
layer, acting as a 'spanning level', installs full independence of applications
(and service offerings) from the underlying network substrates—so, network
configuration is absolutely transparent to customers (Clark, 1988;
Messerschmitt, 1996) The later architectural specification calls for open
protocols of communication between lower and higher layers5. While 'open'
is often mooted in standards discussions, the real meaning is quite simple: the
functionality and specifics of a digital interface are all known, transparent and
readily available for implementation. Proprietary code and processes could be
legitimate in an open interface, as long as there is complete and useful
documentation.6
As is obvious, the emerging cognitive model is to separate, within a
multi-layer system, service offerings and applications from information
facilities7 and therefore to permit: (i) customer independence from the
network provider; and (ii) different but compatible innovation paths at each
part of the system—that is, technical diversity at all levels commensurate
with each level's evolutionary technological pathway. At the same time, this
architecture, building on the definition of open and ubiquitous protocols,
permit seamless interoperability between a variety of network conditions
(local, wide-area, public and private) to be succesful in the marketplace.
An Evolutionary Framework for Infrastructure Standardization
While the technical aspects of successful open standards, such as those of the
Internet Engineering Task Force (IETF), are progressively specified by a
' It is worthwhile to note that the role of OSI protocols include prevision for a layer which probably
could play the role of a 'spanning level' (in the technical jargon it is the ODI 'internet sublayer'). However,
the CCITRT (International Consultative Committee on Telegraphy and Telephony), which is the main
telecommunications authority for standards, has not followed OSI in adopting such a specification and
including it currently implemented network architectures (Quarterman and Wilhelm, 1993).
4
For a general view of the various (conflicting) ways of 'openness' in the context of future
communications networks, see Band (1993)- '
' For a more extended analysis of this point, see NREN (1994), Nil 2000 (1996) and below.
1100 ~

The Internet: a Paradigmatic Rupture in Cumulative Telecom Evolution
user-centric trial-and-error approach, little is understood about the economic
significance of this method as compared with the older method of
carrier-directed, user-passive teleological implementation. Mansell (1994) is
right when she detects 'design configurations' (citing Metcalfe and Reeve,
1990) as the fundamental units around which discussion of the economics of
technical change should be organized. However, the dictats of the
configuration work differently if the user is involved from the outset instead
of the user needs being anticipated by the communications provider.8 This
particular point strongly supports an evolutionary perspective for the
infrastructure standards.9
The Standard Implementation as an Adaptation Process. Genera l ly
speaking, design configurations are embodied in particular standards. In the
infrastructure industries, standards has always been considered as a strategic
arm, setting organizational models with strong 'lock-in' properties. In the
early times of telephone, AT&T, under the Bell System model, imposed and
maintained its hegemony over the US telephone network by using licencing
and franchising mechanisms and strategic manipulation of standards and
intellectual property (Solomon, 1990). Doing so, AT&T succeeded in
imposing a particular technical configuration on the US telephone network,
including, through the control of interconnection, parts that were not under
its direct control.
The process by which infrastructure standards are established is now more
complex than in the early telephone days, but it still remains an area of
competitive strategy (Besen and Saloner, 1988). Perhaps what is radically new
is that the effective emergence of a standard depends more and more on the
potential adopters' attitude—in the same way that innovation improves and
develops considerably during its diffusion process (Silverberg et al., 1988).
Putting it in another way, as the interactions between users and producers
play a critical role in the technical development process (Von Hippel, 1976),
the implementation process of a standard determines its effectiveness.
Standards-setting appears as an evolutionary process in which adaptation
makes the difference and not the initially better designed standard.
* On the Internet, the provider was basically ignored until recently, while the user community designed
their own overlay data communication system responding directly to their immediate and future needs;
the result in traffic growth and innovative applications has been phenomenal, averaging 40% annually
worldwide.
' Other scholars (Metcalfe and Miles, 1994) propose a model similar to ours, defining selection and
variety roles during the standardization process. Both approaches are based on evolutionary concepts of
technical change (Nelson and Winter, 1982; Dosi, 1988; Nelson, 1994).
1101