Voice
Communication in Business Volume 2
Essays on telecommunications,
1981-2002
In the
last 3 or 4 years of the old Millennium, the Internet became the
main topic of interest in communications. Although its main function
turned out the be the pervasion of porno, often to children, it had
even higher ambitions. In addition to its actual purpose as a medium
for data communication, it wanted to take over TV distribution from
broadcast, satellite and cable, and point-to-point voice from the
various long distance carriers or IXCs. Because long distance
charges had fallen to something on the order of five cents a minute
via the IXCs, most of them were in deep financial trouble.
The
Internet, which had never bothered with earning a living and which
had produced a cataclysmic stock market crash in 2001, moved ahead
on its way to taking over long distance.
I
noted some misgivings in the July, 2001,
Business Communications Review.
Lessons For The VoIP
Revolution
Business Communications Review, 2001)
History is
bunk. — Henry Ford
Those who
cannot remember the past are condemned to repeat it. —
George Santayana
It is obviously possible
to send voice over Internet Protocol (VoIP). Further, Parkinson's
Law (data expands to fill the bandwidth available) assures us that
if you build a network and give it away, users will come in droves.
Thus the frequently noted growth-rate for Internet data, vastly
greater than the growth-rate for long-distance telephone calls,
implies that, sooner rather than later, data will far exceed voice,
and voice can be carried as an insignificant portion of Internet
traffic, making a circuit-switched voice network unnecessary.
Note that the
assumptions in the above do not involve paid traffic. Even at
today's rates, long-distance phone calls doubtless generate far more
revenue than data transport over the Internet. If people had to pay
for Internet use in any way proportional to the actual cost of
providing the service, it is doubtful that many would transmit
multi-megabyte files consisting of pictures of ducks, or adorn their
second-hand e-mail with extensive unrelated clip-art. Freedom of the
market place is supposed to let us select what we want by voting
with our bucks. When we simply accept whatever charity others decide
to bestow on us, we may or may not get what we want, and we may lose
what we are getting when the donors choose to no longer donate. Thus
the impending VoIP revolution may, at the very least, turn out to be
a mixed blessing.
There have been many
earlier revolutions in the telephone industry. For local calls,
automatic switching started to replace operators before the turn of
the last century. Translation made it possible, starting in the
1920s, for many switches in large metropolitan areas to act as a
single automatic system. Common control in the 1930s, 40s and 50s,
reduced control costs, increased certain types of flexibility and
speeded service. Then, computer- based stored program control in the
1960s increased by several orders of magnitude what
electromechanical common control had been doing.
In the long-distance
area, microwave expanded transmission capabilities and dropped costs
in the 1940s. Automatic message accounting replaced most toll
operators prior to 1960. Digital toll switching with the 4ESS in the
1970s made possible much larger switches which simplified routing
into major centers. And finally, fiber optics, from the early 1980s
on, again expanded transmission by making 1960s T-Carrier as
practical for long distance as it had been for short-haul trunks,
and dropped costs even more by emphasizing the importance of
switching in the digital transmission format.
All these revolutions
differed from the VoIP Revolution in two very specific ways. First,
they were designed by profit-making, tax-paying companies to reduce
costs and improve service. Long distance service improved
continually, and its cost, in current as well as constant dollars,
went down. It is not clear what will happen to VoIP if companies
offering it suddenly have to earn a profit.
The other major
difference is that the previous revolutions in telecommunications
were conducted under a system of vertical integration. AT&T managed
the overall system; Bell Labs designed the equipment, Western
Electric manufactured it, and the local operating companies and AT&T
Long-Lines used it. As a result, planning could be carried out in
terms of overall system needs, manufacturing could be conducted at
optimum rates, and what others much later called life cycle costing,
could be applied to minimize overall costs and not just the purchase
price of selected pieces of hardware. The cost reduction obtained by
such an approach was considerable. As the Operations Research people
used to emphasize, a global minimum is usually a lot better than a
number of lesser local minimums.
Now that Lucent (the
former Bell Labs and Western Electric) is independent of AT&T and
everybody else, and is just one of many companies like Nortel and
Cisco competing for the equipment bucks of Internet entrepreneurs,
is it not reasonable to expect these manufacturers to push their
latest toys just as Intel and Microsoft do, whether the service
providers need them or not? Backward compatibility and legacy
systems will vanish from telephony as they have from the PC world,
and companies that try to offer Internet service will find
themselves struggling, like PC customers, to keep up with buying
hardware and software that simply will not stay bought.
The need for
specifications and contracts was greatly minimized under vertical
integration. I have often shocked the naive by pointing out that
there is no legal document that can make a system work. All any
legal document can do is fix the blame when the system bombs. When
one company plans, designs, builds and owns the equipment which is
used to render a service for which the customer pays, the motivation
to make the whole system work correctly from the beginning is very
strong: You can't earn money on phone calls the customer can't make.
With vertical integration gone, future business practice will
doubtless focus far more on fixing the blame than fixing the
problem.
The Internet As Camel
The Internet is like the
proverbial camel, a horse designed by committee. It is a
collectivist communication project constructed by self-styled rugged
individualists, first at government expense and then at the expense
of lemming-like venture capitalists who apparently never considered
the financial returns, if any, such a network might generate.
As such, it reflects
more the artistic visions of its creators than the needs of those
who turned out to be its users. Even before the Dot-Com crash of the
past winter, the irrational exuberance of Internet economics was
openly questioned by those of more conventional turn of mind. It
would have been nice if a soft landing could have been arranged, but
when the victims said they knew what they were doing, and it's
different this time, offers of help sounded like sour grapes.
But believe it or not,
it was not different this time, and we have all had a chance to see,
first hand, another example of the kind of boom and bust that seems
to come naturally when people are freed of their need to apply
intelligence to their passions. We may hope, however, that a lot of
the infrastructure can be preserved and put to good use as we add
bits of wisdom to our considerable technical knowledge.
After thirty years or so
of cut and try, the Internet works pretty well at doing what it was
designed to do: letting time-sharing terminals of one main-frame
computer access other main-frames in a network. Mainframes accessed
the original ARPAnet via IMPs, or Interface Message Processors, the
parents of today's routers. TIPs (Terminal Interface Processors), a
variation of IMPs, allowed orphan terminals to enter the network
directly. This structure continues to allow today's PCs, called
hosts (apparently a reminder of the time when large computers hosted
dumb terminals) to access remote data bases.
Most of the time, active
Internet users are neither sending nor receiving information;
instead, they are looking at or listening to information that has
been downloaded, or following instructions already stored in their
PCs. When they send, each transmission is little more than a couple
of key strokes or mouse clicks ... very short bursts of data
generated at 100 b/s or less, although the current standard rate of
transmission is something more than 30 Kb/s, while it lasts.
Downloads contain a lot more data, often megabytes, but come in
occasional bursts at speeds essentially unrelated to what is needed
for presentation to a human. Often, vast quantities of information
are stored on the user's disk drive, and fed out to the system
display, audio or video, at a rate controlled by the user's
terminal. Delay through the system is relatively unimportant, as in
TV broadcasting. When you read a file, listen to music or watch a
ballgame, you don't really care if you see or hear the material in
real time or two or three seconds late.
This illustrates the
principal differences that VoIP will encounter compared to data over
the Internet: Customers will generate about the same amount of
information in each direction, and far more than a few mouse clicks;
further, delay will becomes critical. Anything resembling
Picturephone (interactive video) will simply increase by several
orders of magnitude the amount of information flowing, now
continuously but perhaps at a variable rate, in each direction, with
the added problem of synchronizing audio and video. Because video
compression takes longer than audio compression, but neither can
exceed what the human nervous system has evolved to expect, the
Internet is not the ideal vehicle for two-way real-time human
interaction.
Already there is work
toward making the Internet hierarchical to minimize the store and
forward delay (euphemistically called QOS for Quality of Service)
which data packets shrug off as they wander all over hell's
half-acre going from Philadelphia to San Francisco. More nearly
direct routes (that is, those with fewer intermediate nodes) will
reduce transit delay, but will not help in any way to reduce the
delay necessary to acquire enough speech to fill even a
reasonable-sized packet.
Already, it appears that
packet headers for VoIP will be much longer than their voice
pay-load, and when IPv6 thrusts itself on the VoIP world, whether or
not designers, operators or customers admit they are out of
addresses, IP headers will get even longer. This will reduce even
more the efficiency of VoIP. Designers, faked out by the vast
availability of overbuilt fiber-optic routes, will be blindsided by
this just as the telephone industry was by the need for new area
codes brought on by pagers, cell phones, residential data lines,
etc. I did not ask to move from 609 to 856; it just happened.
Economics To The Rescue
About the only thing
that can slow the coming of VoIP is--surprise!--economics. The glut
of optical fiber, in addition to making today's Internet possible,
has also lowered the cost of regular circuit-switched voice
telephone calls. Today, anybody who is paying more than about 8
cents a minute for interstate long distance simply likes to give
money away, and those who look can find much better bargains.
Out of this small
charge, the present long-distance interexchange carriers (IXCs) pay
access fees to the local telephone companies at each end of a
connection for gathering up the long distance traffic in the first
place, delivering the traffic in the second place and, finally, for
providing the basic information needed for billing. The Internet
Service Providers (ISPs) do not pay access fees. They are not only
subsidized by the government, but also by the local telcos and the
long-distance carriers who raise the revenue that pays for local
customer access.
Of course, there is
nothing wrong with subsidies from the government or anybody else.
The railroads, the Pony Express, highways, air mail, air travel
itself and many other worth-while endeavors got their start and many
continue by riding on subsidies.
But a certain level of
fairness is needed: the so-called level playing field. It appears
that the only way the Internuts can undercut the price of regular
long distance is by not paying access fees for VoIP, just as they
don't pay them for the data they carry. If they have to pay at the
same rate that the IXCs do, or if the IXCs can make a case that they
shouldn't pay access fees either, it is hard to see how the
Internuts can beat IXC prices.
So, maybe we just might
want to see how some of the current economic issues such as the need
to make a profit, newly discovered by the Internuts, play out. Maybe
VoIP will still be able to undercut the existing telephone network,
and VoIP COs and PBXs will be a lot less expensive than traditional
ones, even though their designers don't seem to have a clue as to
what features and services are expected by customers. I, for one,
will be most interested in seeing a router that can handle 50,000 or
more simultaneous voice calls in its spare time while it is also
handling all the data that presumably will vastly transcend voice in
quantity if not quality. I want to see how the Internuts will handle
charging and billing, and dividing up VoIP profits among their
various networks. I want to see how the problem of delay is actually
solved for voice connections.
Conclusion
The 1975 telephone
revolution was not, as some say, from analog to digital. It was from
electromechanical to electronic switching. About half the electronic
PBXs were analog, and of those which were digital, most were not
compatible with T-carrier, the digital transmission system that gave
meaning to being digital. In this world, Mitel made its fortune by
not being digital, by switching the analog signal in the form it was
received from conventional phones. Why spend money for a codec in
each end of an intra-PBX connection less than two feet long? In the
CO field, where very large local switches (up to 100,000 analog
lines, plus trunks almost 100% digital) were needed, digital
switching eventually could make an economic case, although 5ESS used
a space- division concentrator when dealing with lines to
conventional telephones, modems, fax machines, etc. If ISDN had come
in as of 1985 or so, being digital all the way to the customer at 64
Kb/s could surely have pulverized the 300 b/s modems of the day.
With 1ESS 15 years ago,
conversion to digital on the trunk side was essentially cost-free;
today, conversion to VoIP, on the trunk side of a PBX or local CO,
still has to be far less expensive than providing proprietary VoIP
phones, voice compression, echo cancelers, etc., on a per-line
basis, just to use Internet technology. As many have noted, few
customers will demand that a working PBX, all paid for, be torn out
so that another PBX, not even bothering to offer standard features
but advertised as state of the art, can replace it. The question in
1980 was, Why should I buy a digital PBX when it won't let my
secretary screen my calls? Today, one might question spending money
for a VoIP PBX for similar reasons.
There is, however,
something new in today's mix: digital cell phones. Their coming en
masse in the last few years, partially a result of $5 local calls
via competitive pay-phones, is nothing short of phenomenal. If they
can ever agree on one standard digital approach for both voice and
data, we would have economic and technological justification for IP
COs, PBXs, and VoIP-based networks. We would have customer-provided
signals already tailored for end-to-end Internet transmission. That
could very well make VoIP triumphant.
[
Top ] [
Table of Contents
] |
