IP: The World according to Farber -- notes on a talk presented in Japan in Dec 2000

[Dave Farber <farber () cis upenn edu>]

The following are notes produced by Adam Peake of GLOCOM of a talk sponsored by
GLOCOM for senior Japanese persons during my visit to Tokyo in December 
2000. Dave Farber


The Future of Telecommunications and Networking--Predicting the Unpredictable


Professor David J. Farber
Chief Technologist, US Federal Communications Commission
Alfred Fitler Moore Professor of Telecommunication Systems, University of
Pennsylvania
Visiting Professor, GLOCOM


December 6, 2000, Tokyo
Notes - Adam Peake



The merging of computers and communications.


-Historical observation

The Internet began with the idea of hooking computers together, a network
diagram with large computers and small cloud that was the network. Now,
basically no one cares about computers, it's the network and its ability to
tie people together and people to information that is the profound change
in our attitude toward networking.


-Some predictions

The network space is at a significant crossroads. A US perspective, but we
can expect a number of possible outcomes. A world like TV, where people
have limited availability of programming, viewing shopping and interacting
at that level. And there are players trying to push the net in that
direction from the user viewpoint.


Or we cold see a more interesting space, a Stephenson ("Snow Crash", Neal
Stephenson) like world of virtual spaces, or evolve in directions we don't
understand.


Many of the decisions we are taking now will effect this future and we must
try and understand what the implications may be of these decisions. Like it
or not, governments will profound impact on this direction.


The commercial world is undergoing a series of experiments, some successful
and not, but we are changing customs and long established practices. A
problem is that we tend to believe that whatever we see as exciting at the
moment is what the future will be. Caution that we're experimenting. No
reason to believe that the current direction will be profitable
(materially/non-materially), and we can see this in technology and in
applications. We believe today the web is a great thing, suspect in 3 years
time we will look back and wonder why we ever thought that. It's an ongoing
experiment.


Technology forecasting - dangerous territory, one where we have a poor
record. Certainly didn't anticipate the net becoming what it is today, or
would be as rich as B. Gates... but a few remarks:


-Things driving network technology

The availability of very large data stores means we can do things in the
business world previously unavailable, e.g. models that check backward for
validity and forward for speculation. The potential new generation of image
capture and display systems: "reality" TV, very high resolution images with
the ability to capture them, providing the feeling of immersion on a flat
panel. Aided by a better understanding of how the eye works.


The greatest change will come from the changing structure of the
communications system. For the last century we have been dealing with
electrons. We have optics now, but they get converted to electrons to get
routed and switched. The coming decade will see the introduction of all
optical networks. First in the wide area network, then down to the local
network to all optical networks with light in/light out. It will happen,
the question is how and what the economics are. 200 waves at 60/80 gigabits
per strand. And many strands in a cable and many cables. Potential serious
economic impact, we haven't many successful cases where the price goes to
zero and we stay in business!


-Implications on the technical infrastructure

Unreasonable to expect we will be able to do the kind of packet routing we
do on the current net. Protocols like IP will probably not be useable in
the way we think of them today. IP was built for a different regime of
networking, may still be called IP, retain important philosophical
considerations, but not be the same.


High bandwidth network applications will require us to pay more attention
to the rest of the computer environment. With a struggle we have been able
to get gigabit level information into our computers, but they weren't
designed for it. Absorbing any significant percentage of a 60 gigabit
stream into a computer system is daunting. Problems of current software
will be even more daunting, too much bulk and can't handle the modest data
speeds of today, just not realistic to expect them to work in the all
optical network. Assuming all optical networks will happen, we will have to
rethink everything we currently do. This happens periodically in computing,
the last time probably being when the microcomputer was introduced. We're
beginning a time of transition when once successful companies can be
expected to fail and new players emerge.


Looks like the first national optical link will come  in around
2003/4, it will rapidly go into metro areas (examples of which we see
already in Milan and soon to be in Chicago.) Shortly after a distribution
system will come in to service those MANs and soon after LAN technology
based on optical technology.


When we began gigabit network trials people asked why when there's no need
for such bandwidth. No doubt we will hear the same question of 60 gigabit
links. But expect markets to emerge. At a recent talk, new business school
students responded to discussion of a future with significant bandwidth not
with questions about how much will it cost, but when can we get it. We are
approaching a significant discontinuity in technology, we have little time
to understand what its implications are.


-Politics of communications, immersion inside the Washington beltway

A world populated by lawyers, economists and lobbyists. Technology in
Washington is like a mushroom farm, grown in the dark, fed manure.
Interesting to observe that the rest of the world has a habit of following
the US 5 years later, just as the US has recognized the problems and is
trying to fix them.


Will the US escape regulation? Odds are small that it will. Currently no
one wants to regulate the Internet, for the good reason that they don't
understand it. But the vulnerabilities of the current network may bring
regulation: it is relatively unreliable and lacking any decent notion of
privacy protection. These problems may bring in government. If the network
melts down for a while, to the extent that large parts of business stops,
then government will react. When it happens --pretty good chance it will--
Congress will pass laws to make sure it doesn't happen again. Probably the
worst thing Congress could do, but a reaction that will be unstoppable.
There are a number of vulnerabilities that we haven't paid attention to
that we must now.


The Internet has challenged a number of powerful companies. Example, the
local exchange carriers and other legacy telecom companies. On the one hand
they say don't regulate let the market determine, but when it applies to
others, they ask to regulate "them." Pessimistic that government will not
end up regulating and caution it will introduce problems.


-Policy issues, not just technology issues

In the US, the most likely way broadband will get to the home is over the
cable system. Currently a choice between cable or DSL. Put politely, Telcos
have been slow in delivering DSL, the cable operators have been more
aggressive and in light of recent/ongoing merger developments there's
reason to believe they will be more aggressive in the future. But a key
policy issue arises because cable operators control program material with
very little restriction that they have to show anything. There are must
carry rules that if you broadcast television the cable company must carry
it in the local environment. But that and some small ownership constraints
are the only limitations. So leads to the question will the broadband
system be an open system or one carefully controlled by the cable operator?


Relates to the question whether the Internet in the future will be more
like television or like the chaotic Internet of today?  Will you have to
deal with a cable operator to make sure you don't compete with their
programming? Will they constrain what you do can do on the system and the
material you can transmit? A major policy debate in Washington and one that
will shape the way the US goes in the consumer Internet space. If the only
viable broadband delivery system is closed we can expect the net to evolve
into a very dull structure, if it's open it may be interesting.


Expect the issue in different contexts will arise around the world as
communications companies deregulate and there is a need to make them open.


-Mobility and Security

We are about to switch to a next generation mobile cellular system. It
seems to offer good data rates --at unstated prices-- but it lacks
spectrum. The US, and suspect other countries, have made a mess of spectrum
allocation by treating spectrum as fenced property, and wasted spectrum by
giving it to people who don't know what to do with it and have no useful
need for it. So the situation now in the US is we don't have the spectrum
available for any rational deployment of 3G.


Expect 3G to be a relatively short lifetime system. Expect we will have to
move to software defined radios to get around the spectrum
scarcity/unavailability problem. Expect we will change our understanding of
the rights of the spectrum "owner" with the emergence of a notion of
spectrum guardian rather than ownership.  The property laws of Scandinavia,
e.g. I can access your land of I do no harm, could apply well to spectrum.
We have technology to do it and will probably have to do it sometime around
the middle of this decade. Current waste of spectrum is defeating the
rational use of mobility.


-Local wireless, Bluetooth and equivalents

Communications between control devices. Many potential opportunities, but
potential problems in that they use unlicensed spectrum, with no rules of
conduct in a crowded space. Expect rules of conduct in the unlicensed band.
Not licensing but rules of behavior. A bill of rights and obligations.


-Security

The world of the always-connected citizen, in their connected houses,
raises serious security concerns. We will be increasingly vulnerable to the
bad design elements of our current infrastructure.


Difficult to add security after the fact. We are interconnecting the
Internet and telephone system, usually via gateways to the SS7 switching
system, and they were not built to have people outside the carriers
interconnecting. But anyone with a relatively easy to obtain CLEC license
has a right to interconnect. And while offering great benefits, it also
presents the potential for trouble. The Internet is asking things of the
system outside of what it was designed for. We need to allow people to
interconnect but also be aware of the potential risks


-Technology and society

Will ecommerce world really blossom: not decided, there are questions
whether consumers will actually use it with vigor. Questions about some of
the delivery systems, that the operator should receive a cut of anything
bought over their conduit. A greater influence will be taxes and how to
collect them, particularly in the US and Europe. So we have nation states
wanting their taxes, carriers wanting a cut from what used to be a
relatively passive communication channel, and consumers who haven't
necessarily decided if it's a good thing to begin with. Then add the
security problem.


-Potential problems that need attention
Cultural differences between countries, e.g. the purchase of Mein Kampf and
tension between nations and vendors. Attempts to apply the rules of one
country onto the citizens of another. Risk of the most conservative
countries dictating acceptable content on the net. Not enough attention
being paid by corporate interests that may be severely effected by this.


Conflict between the freedom of the network to carry everything and freedom
of government to understand what's there for reasons of law enforcement and
security. If one trusts government it is not an issue, but around the world
there's increasing mistrust of government. We have technology that enables
us to watch almost any and all communication. The clash between government
concerns and the rights of individuals will be major battleground. In a
sense it impacts the business sector, as our ability to deploy new
technology depends on peoples willingness to adopt it and government to let
us. Example in that there are many benefits to deploying cryptography
besides privacy, but governments have stopped us from deploying it and so
made the networks less interesting. Business people must take an active
interest in understanding the issues and at least come down on one side or
another.


We want and hope for technology to create the virtual home or lifestyle,
but many are concerned that government then will not only be able to invade
my external email, but also be "in" my home hearing what I say to my
friends and family, knowing where I am and what I am doing. Technology is
offering great opportunities for the future, but this darkside could stop
us from getting there.


There's a tendency toward shortsightedness in research and vision, the
downsizing of research groups, asking them to only look 2 or 3 years ahead.
Like an army without advanced scouts, you can expect to walk into trouble.
We are going to experience some very significant changes, and the lack of
anybody in our industrial giants looking into the future could cause
serious problems through lack of awareness of what might be coming. We have
a responsibility to think and worry about the future.




Q&A


Q. You mention that control of Internet content by providers would likely
create a dull world. But even with today's limited bandwidth there is
interesting chaos. What is your concern?


A. The cable operator is under no obligation to offer any access to the
digital channel other than what they choose to offer. Hypothetically, they
could refuse to allow any ISP other than some specified company onto that
digital channel. The Federal Trade Commission would react to that, and
would probably force them to do otherwise.  But looking at the competitive
future we also have to consider the ability to tie the data channel into
the MPEG4 streams. There is no obligation for the cable operator to allow
anyone at that MPEG4 stream except those companies that broadcast. The
cable operator may have a preferred ISP who they also give access to the
MPEG4 channels, that ISP would then be able to offer spectacular service.
And within that preferred ISPs world it could control what you see and
perhaps cut you off from stuff on the general Internet. No legal
requirement that they wouldn't make the experience very limited/dull.
Whether they would or not is a different question, but suspect they might
try. If broadband is to be interesting it will require access to high
quality video streaming and that's where the lock is in the US.


Q. Since the progress of all optical communications is so rapid and even IP
may not be used in the future, do you mean that a shift to IPV6 that the
Japan government is going to promote will be enough, or to some even more
profound change.


A. Given the nature of what can be done on an all optical network it
appears that IP is too elaborate a protocol, designed for a totally
different speed, latency and product. Difficult to support high data rates
because of this complexity, but it was designed in an era when bandwidth
was in short supply and error rates high, and neither are the case in all
optical networks. Also in high performance systems we are trying to replace
IP stacks with application driven control. Things move up to the
application level. IP does not make sense in an optical network except as
philosophy.


Switching to IPV6 gives address space, but we are distributing it without a
view to the future so may be creating a new scarcity by this careless
approach. IPV6 doesn't solve many issues, and not clear it will deploy in
any meaningful way soon in the US. Could deploy tomorrow, all it would take
is for Microsoft to included it as the primary stack. Routers can handle it.


IPV6 doesn't contribute to stability, but it's a new protocol with new
software and we don't need something new that might cause further
instability even unintentionally. Some have converted, but a national
conversion may be difficult and worry about stability, we have enough
trouble with software as is. While it has advantages, IPV6 is not
propelling you into a brand new environment.


Q. What major pieces need to be put in place to make an all optical network
feasible?


A. Evolutional approach. First with IP run over it, the core optical first
with electronic switches up to the core edge, a protocol conversion at the
edge. But it will rapidly spill out as the economics of the conversion take
effect.


Q. Explain more about software radios


A. Electronics replaced by software. Simplistically, software takes the
info from the antenna and handles it digitally as bitstream. Frequency and
protocols are all defined in software and are changeable as software is
changeable. Adapting to the environment it's in through software,
"sniffing" the air for optimum service, etc. It is the way to realize
rational spectrum use.


1992 report talking about spectrum use written by Paul Baran is the basis
for much of the modern thinking on spectrum management.


Q. Rational of IPV6 is that we are facing a shortage of address space,
Japanese PM Mori's recent speech mentioned that countries other than US,
Europe, Australia and Japan are facing an address shortage, and then we
have the needs of mobile, etc.


A. Some bad judgements resulted in bad IPV4 address space allocation.
Worried that we are making the same bad judgements with IPV6 allocation,
with poor control and waste. The new software phenomenon may cause a lot of
new problems; let industry do the changeover first, see if it's stable then
deploy to the public. And there's breathing space in the current system,
and while uncomfortable with NATs (and everyone is) and other solutions and
they work.


Q. About intelligence activities of nation states. Will the massive data
flows on the net today hinder their searching? It seems that some of the
problems we see from the intelligence community come from them looking for
a job in the new world.


A. Don't believe they have any problem in analyzing what's out there, don't
underestimate their capabilities.


Most US domestic wire tap activities are conducted by the FBI. Difficult to
understand what their threat model is, and whatever it is is classified.


Two problems.  First, intrusion, that they can look at anything in ways
that the physical world would not allow. Second, their tools are pretty
bad. Have to be worried their software might cause trouble in the net.
Carnivore seems terrible in this regard.


Something to consider. Given that email is trivial to forge, trivial to
intercept, and sent unencrypted, why haven't we got good encryption in our
email systems?  It's not technically difficult and there's a market for it.
So what's the reason we don't have it?


The UK is much worse, their law enforcement seem to have some god given
right to look at anything they want. Fortunately the US is far more
specific.


Q. Mentioned that government is still trying to understand the Internet,
what do you think might cause them to begin regulation?


A. Most likely cause would be a serious failure. The other would be that
industry in the US often tries to use regulation for a competitive edge.
It's not unusual for a company to ask for regulatory relief in some area,
but also ask that some area that impacts their competitor but not them
should be regulated. This may influence some regulation to be put in place.


We have another problem in that it takes about 3 years for any regulation
to be put in place, and in Internet time 3 years can be 2 generations
behind.


Government knows the words for the Internet and they have plenty of
technical advice, but not people who understand technology working with
decision-makers. The cure is to bring in more technology people to work
more closely with government.


Q. 3G mobile may have a short life time because of a shortage of spectrum,
this may be true for the US, but other countries have enough.


A. Depends on how short "short" is. Will be more difficult in countries
where there is stronger competition. Major US cities have 7 operators and
that will be interesting when a single new standard is introduced.
Operators in very competitive markets may not have the revenues and high
profits from existing services needed for investment in new 3G networks.
DoCoMo's perspective is no generation will last more than 10 years, 3G will
probably go less. DoCoMo has about 20% of their R&D manpower on 4G already.


3G doesn't have multicast -- strange in a bandwidth limited system. It's a
telephone oriented system, a telephone system carrying data. Give it 8
years. Still, no one seems to have any idea how much new 3G visual services
will cost. Would much rather go the other direction and build a mobile data
system and run voice on top.


Q. What's your opinion on frequency auctions - a non-FCC view.


A. The results of some EU auctions make it difficult to see how they will
ever be able to repay. Expect some operators will be going back to their
governments in a few years and asking for a discount. Auctions also give
the notion that the spectrum is owned, and that's a bad idea. No
alternative to offer, but there must be a better way.


Not sure that US cellphone operators have the resources for high auction
prices, competition is tough, prices low with corresponding low profit
margins. 7 cents/minutes, anywhere.


Another problem with regulation in the technical environment is that by the
time you can go after a company for some market failure, it can be very
difficult to recover from that failure.


Q. What might we expect from the next administration?


A. [comments before Supreme Court ruled.] Don't believe either really
"understands" the Internet but Presidents don't get
involved at that level. At the FCC, Commissioner Powell looks like the next
chairman, he a centrist republican who will let the free market run a
little bit longer than current Chairman Kennard.


Justice department. Always the expectation that republicans will be soft on
monopolies, but they never have been.


Guess that with the split between the parties we will see lukewarm
centralist politics. No new communications act, no strengthening of law
enforcement, etc. And a peaceful time may be the best thing right now.


END



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