Accessibility
"Open Access" is Just the Tip of the Iceberg
by Jerome H. Saltzer
October 22, 1999

The argument between cable companies and municipal regulators that has
been labeled "Open Access" is actually just the first of a series of
arguments destined to surface as it begins to dawn on customers that
cable companies stand not only as access providers but also as
gatekekepers to the Internet.

Supporters of "Open Access" mean to forbid one, somewhat indirect form
of gatekeeping:  service bundling.  Access providers would like to
bundle access to the Internet with a collection of other services,
such as e-mailboxes, internet name lookup service and web page
hosting.  For many customers, this bundle is exactly what they want
and need, and bundling is a convenience.  But there are other
providers of those services, some of which may offer a more attractive
deal, with lower price or better features than the outfit that happens
to have the wire franchise.  And a more ambitious customer can even
provide these services himself.

But reports are starting to come in that cable companies are also
exerting several other forms of gatekeeping.  These other forms have
three features in common:

    - The majority of customers don't notice them, but they interfere
with innovative Internet use by leading-edge customers.  When only a
small group is affected, it is hard to amass the necessary clout to
repair the problem.  But leading-edge customers are important, because
they represent the future of the Internet.

    - The provider can offer a technical excuse that control is
necessary to ensure good service.  For more than sixty years, the Bell
Telephone System offered exactly this argument:  control is necessary
to protect the network.  Unfortunately, control can stifle innovation.
It did so for telephone service, and it promises to do so for Internet
service.

    - The control operates to the financial benefit of the provider,
reducing incentives to overcome the technical excuse.  Thus control
becomes anti-competitive.

Here are five examples of gatekeeping that have been reported by
Internet customers of cable companies.  Customers of Digital
Subscriber Line (DSL) service have reported the same or similar
examples of control:

1.  Video limits.  Some access providers limit the number of minutes
that a customer may use a "streaming video" connection.  Today,
streaming video is not widely used, because it provides movies that
are small and erratic, but one day streaming video is likely to become
an effective way to watch television programs from many
sources--chosen by the customer, not the cable company--or to purchase
pay-per-view movies.  The technical excuse for this restriction is
that the provider doesn't have enough capacity for all customers to
use streaming video at the same time.  But cable companies have a
conflict of interest--they are restricting a service that will someday
directly compete with Cable TV.

2.  Server restrictions.  While advertising the benefits of being
"always on" the Internet, some providers impose an "acceptable use"
contract that forbids customers from operating an Internet service,
such as a web site.  The technical excuse is that web sites tend to
attract lots of traffic, and the provider doesn't have enough
capacity.  But again the access provider has a conflict of interest,
because it also offers a web site hosting service.  The bundled
version of this service satisfies many customers, but a customer with
only a mildly ambitious web site will exceed the parameters of the
bundled service and fees for extra storage space and high traffic
volumes add up rapidly.  (Some providers have adopted a more subtle
approach: they refuse to assign a stable Internet address to home
computers, thereby making it hard for the customer to offer an
Internet service that others can reliably find.  And some access
providers have placed an artificial bottleneck on outbound data rate,
to discourage people from running Internet services.)

3.  Fixed backbone choice.  Access providers choose where they attach
to a long distance carrier for the Internet, known as a "backbone
provider".  The route to the backbone provider and the choice of the
backbone provider are important decisions, bundled with the access
service.  If you reside in Massachusetts, and you connect to a
computer in your office in the next town, unless your office uses the
same access provider, your traffic may flow from Massachusetts down to
Virginia and back.  This detour introduces delays, which can
significantly interfere with some kinds of service, such as video
conferencing with your boss or interactive file editing.  In addition
to distance-related delays, you may encounter distant,
response-slowing congestion, or even inability to communicate with
your office when a hurricane hits Virginia.  Your access provider
again has a conflict of interest--attaching to the nearest, most
effective backbone provider might divert revenue from a backbone
company in which your access provider has a financial interest or
other business dealings.  More important for the future of innovative
services, if a new backbone provider offers a specially-configured
low-delay forwarding service which is just what is needed to carry
telephone calls over the Internet, your access provider (which may
also offer telephone service) may choose not to connect to that new
backbone, effectively preventing you from using a better service.

4.  Filtering.  Data is carried on the Internet in batches called
packets, and every internet packet contains an identifier that gives a
rough indication of what this packet is for:  e-mail, a web page, a
name lookup, a remote login, or file sharing.  Several access
providers have begun to examine every packet that they carry, and
discard those with certain purposes, particularly those used for file
sharing.  The technical excuse for this filtering is that many users
don't realize that their computer allows sharing of files, and
filtering prevents other customers from misusing that feature.  But
some access providers have imposed filtering on every customer,
including those who want to share files.  There is a similar risk that
pressures to restrict access by children to undesirable content such
as pornography may lead an access provider to impose content filters
on all of its customers, including those who disagree with the
particular content restrictions.  And again, there can be a conflict
of interest--the access provider has an incentive to find a technical
or political excuse to filter out services that compete with the
entertainment or Internet services it also offers.

5.  No home network.  An increasing number of homes have two or more
computers interconnected by a home network, and as time goes on we are
likely to find that this home network connects television sets,
household appliances, and many other things.  Some access providers
have suggested that they aren't technically prepared to attach home
networks, but the technology for doing it was developed in the 1970's.
In refusing to attach home networks, providers are actually protecting
their ability to assign the network address of the customer.  By
refusing to carry traffic to internet addresses they didn't assign,
the access provider can prevent the customer from contracting for
simultaneous service with any other Internet access provider.

What is actually at risk in these examples is the very integrity of
the Internet's design.  Since 1975, when the architecture of the
current Internet was laid out, the number of users has increased by
nearly a millionfold; the power of computers has increased by 1000
times while their cost has dropped to one thousandth; the
communications links that make up the network have increased in speed
by a million times; and the Internet is being used in ways completely
undreamed of at the time of its design.  The primary thing that has
made possible this remarkable evolution and adaptation is one simple
design principle, called the End-to-End argument.  The End-to-End
argument says "don't force any service, feature, or restriction on the
customer; his application knows best what features it needs, and
whether or not to provide those features itself."

Each of the five gatekeeping restrictions, as well as the service
bundling that triggered the Open Access debate, is thus at direct odds
with an important architectural principle of the Internet.  the
End-to-End argument is the principle that has enabled the thousands of
innovative applications we see today and it will continue to enable
that kind of innovation only so long as it is not interfered with.

At this point there are at least four possible outcomes:

- Everyone sits on their hands while providers continue with their
present plans, bundling access with unwanted services,
anti-competitive restrictions and lack of customer flexibility.  This
outcome can be expected to lead to stifled competition and innovation
and slower market acceptance of broadband services.

- Local communities can impose licensing regimes for Internet access
that disallow service bundling and that require a minimum of access
constraints and a maximum of customer flexibility.  One hazard with
this outcome is non-uniformity, with every community imposing a
slightly different variation.  It is also an intimidating prospect for
smaller communities, who fear that taking such a position will lead to
a lawsuit from a media company with deep pockets.

- The Federal Communications Commission can impose a uniform
regulatory regime with the same goals.

- Industry members can decide that providing unrestricted local access
not only has benefits to the Internet community as a whole, but by
encouraging innovation and more rapid deployment of high-bandwidth
Internet connections it will actually be beneficial for their own
business.

The first outcome is looking increasingly unlikely, as customers and
cable competitors alike begin to understand better why the Internet
works the way it does and the implications of some of the emerging
practices.  This leaves two regulation-based alternatives and a
non-regulatory one grounded in a larger view of the business
opportunity.  It will be interesting to see which outcome the industry
chooses.

---------------
Mr. Saltzer is a professor of computer science at the
Massachusetts Institute of Technology.