Connectivity woes in Boston

[mike at bilow.com (Michael Bilow)]

The situation is a little more complicated than you paint it, I think.  By
way of disclosure, I should say that I have watched the current mess
evolve as a consultant for a number of ISPs, some of whom are CLECs
(Competitive Local Exchange Carriers, which means they can sell you dial
tone), and I was at one time a leading advocate for ISDN.

* Cable Modems

First, the fundamental problem with cable is the same as it was back when
people started talking about it in the early 1990s: downstream bandwidth
is shared.  Since US television channels are about 6 MHz wide (for reasons
having to do with what could be made out of vacuum tubes using 1940s
technology), and this by coincidence is just about right to carry a 10Mbps
Ethernet signal, the idea was that a cable provider could allocate a few
television channels to data and stick a lot of their customers onto a kind
of virtual Ethernet LAN.  When the cable companies started planning for
these systems, they had little expertise with data and were looking for
existing ISPs as partners, so I got to hear a little of their reasoning.

For example, since at the time the standard modem speed was 28.8kbps, and
the rule of thumb for dial-up modem banks was to have one modem line per
ten customer accounts for residential services, the cable people figured
that a 10Mbps circuit had sufficient capacity to serve 100Mbps using that
same 10:1 ratio.  Never mind that this makes no sense, it was the driving
metric in the early days of cable modems.  Of course, 100/0.0288 would
then give the number of customers who could be served by a single cable
channel, or about 3500.  When I first heard this, I literally started
laughing, and the cable people did not like that.  Even at that point, we
had enough experience using real Ethernet to know that connecting 3500
nodes to a single segment was impossible by a factor of 100.  In office
LANs running at 10Mbps, the rule of thumb is to look at segmenting when
the number of nodes exceeds 30, and you might get up to 100 or so if the
type of LAN traffic is exceedingly light and friendly.

As a result, the first thing the cable companies discovered was that the
whole infrastructure plan upon which they had relied for their business
models was totally off-base.  Based upon market penetration estimates and
computations such as I describe, the plan for one city of about 90,000
people was to use five cable channels!  In practice, by the time 50 or so
households were put onto a channel, customer complaints would be pouring
in.  I was quoted once by a reporter making an analogy to a financial
pyramid scheme: the first people in get lots of bandwidth, they tell all
of their friends and acquaintenances how great it is, and then suddenly
there is no bandwidth left to go around.

In order to salvage cable modem technology, it became necessary to greatly
restrict the scope of utilization of any particular channel.  The concept
is functionally similar to cellular telephones, where bandwidth can be
reused because all of the signals are low-powered and local.  This meant
putting some sort of data infrastructure, essentially a cable router, into
each cable "neighborhood" (which is analogous to a cellular "cell").  
This also removed most of the reasoning for upstream bandwidth
restrictions, but they stayed because the market had already proven that
customers would tolerate them.  In fact, it is easy to buy services such
as Cox @Work, which is basically the same thing as Cox @Home except that
you get a better contract with some service guarantees, have to share your
channel with less people in your neighborhood, and can see higher upstream
bandwidth -- all for about ten times the price.

It is also important to understand that cable modems were not designed for
Internet access in the first place.  The technology largely antedates the
Internet boom, and cable modem rollouts were going experimentally years
before the Internet emerged into public awareness.  Rather, the services
which motivated the design of cable modems were video on demand and home
shopping (VoD&HS).  Once you know this, a lot of the design, such as
having 3500 people all watching the same thing, starts to make more sense.  
The Internet really caught cable providers by surprise, and it was
probably not until 1999 that the last vestiges of the VoD&HS plan died a
quiet death in the industry.  Even today it occasionally wakes from the
dead and staggers, zombie-like, into an FCC hearing.

* ISDN and DSL

Salvage was the goal of DSL from the very start: the whole design is
intended to do something useful with the existing infrastructure of copper
wire telephone loops running out to nearly every house.  The basic
technology of the telephone loop has not changed since Alexander Graham
Bell was alive, despite innovations such as direct dialing, and it is not
uncommon to find 50 year-old instruments still in active service.  On the
other hand, all of that copper wiring that has slowly grown into place
over the last century is now worth tens of billions of dollars, at least
measured by what it would cost to replace, and has rights of way that
would be impossible to obtain in the modern era.  In the vast majority of
cases, however, we are still using these copper loops for analog voice
circuits carried by modulated "battery" current just as we have done for a
century, and we have resisted all attempts to change this.

The natural solution to making more efficient use of the copper loop is to
convert it to a digital signal rather than an analog one.  The main
problem of copper loops is that they get old and noisy, so going digital
allows the use of sophisticated techniques such as error correcting codes
and bandwidth adaptation.  In fact, it is possible to prove that the
channel capacity can be maximized by trading off bandwidth for redundancy.

ISDN was the leading contender to do this using a circuit-switching model,
and it has digital equivalents to all of the usual analog telephone
operations such as dialing and ringing.  The terminology is obtuse --
"ringing" is replaced by "call offering," for example -- but the basic
concept is the same.  ISDN connects you to the telephone system, and once
connected you have full interoperability with tradtional analog services.  
Contrary to popular belief, ISDN is still in wide use: every time you dial
into a modem which can do V.90 56kbps, that answering modem is connected
via ISDN.  What killed residential ISDN was not regulatory obstacles such
as pricing, nor even telephone company incompetence installing it, but the
simple fact that by the time it arrived no one really needed it.

DSL took the opposite approach to the digital loop.  Instead of defining
all of these elaborate equivalents to the analog telephone system needed
for circuit-switched networking, DSL throws in the towel and provides only
packet-switched networking.  The conventional wisdom was that this would
not work for carrying voice circuits (although we now have Voice over IP),
and to a great extent this is true.  In fact, most providers who sell
voice over DSL do so by multiplexing analog voice!  For data, however, the
relative simplicity of DSL is a great advantage.  Since DSL cannot be
circuit-switched, what you get is a kind of leased line to some data
provider, and for economic and regulatory reasons the data provider has to
connect their equipment directly to the end of your copper loop.  This
requires a physical presence at every telephone loop terminus, or at least
at every central office, giving the telephone companies themselves a
natual advantage by requiring their competitors to buy from them and to
depend upon them for all maintenance.

DSL is still quite solid technology and I disagree that it has limited
expansion capability.  It was never designed to have expansion capability,
but just to make the best possible use of 100 year-old infrastructure.  
In this sense, it still has more going for it than any other competing
approach.  The fact that DSL companies are going out of business one after
another has nothing whatsoever to do with the technology.

Several providers are still actively selling DSL locally.  Verizon (Bell
Atlantic/NYNEX), of course, is selling DSL, although their service is
horribly designed and close to useless as a result.  Conversent, a local
CLEC who bought IDS (a RI ISP/CLEC), has a presence at many local central
offices and has not gone the way of the national providers such as Rhythms
and Covad.  (I should say that Covad has a plan to try to rescue the
company and continue operations, but it is doubtful at best.)

My personal opinion is that DSL is too elegant to fail.  It solves a
problem very neatly by making efficient use of available resources.  I do
not see any purported alternative on the horizon, such as metropolitan
area wireless networking, as having nearly as much potential for success.  
The fact that the demand is there was proven during the heyday of the
national DSL providers who are now being driven into bankruptcy.  DSL is
going to have to overcome serious and possibly fatal damage to its
reputation, now that so many people and businesses have been burned in
these bankruptcies, but eventually I think it will be the winning
technology.  The concern is whether there will be any DSL providers left
other than the local monopoly telephone companies.

-- Mike


On 2001-08-10 at 09:37 -0400, Jerry Feldman wrote:

> There is a good reason why cable companies want to restrict the use of 
> servers. For the most part, the cable is set up with a big pipe to your 
> system and a very small pipe from your system out. This is why 
> Mediaone initially set up 1.5Mbps download and 300Kbps up. It has to 
> do with the repeaters and other hardware. This is one reason why the 
> cable companies are sticking to residential internet service. With digital 
> cable and digital phone services increasing, chances are that cable will 
> be available as a business service and that the reverse channel 
> limitations will be removed in the future. 
> 
> There are some economic reasons why they don't want to allow full 2 
> way service. T1 and other services currently sell for significantly more 
> bucks than cable. 
> 
> As far as DSL is concerned, it is a copper technology with limited 
> expansion capability. But, I think the limitations are more economic. The 
> phone companies have been very slow to develop high speed networks to 
> consumers. ISDN virtually died on the vine because the phone 
> companies tarrifed it out of the reach not only for consumers but also for 
> many businesses. The phone companies are using their monopolistic 
> advantage to limit or prevent the loss of their traditional markets. ISPs 
> were registering themselves as phone companies so they could provide 
> DSL to their customers, and the phone companies were very late with 
> joining the DSL bandwagon, and then by using some practices to cause 
> ISPs to lose business. I do think that the regulators will step in. 
> 
> On 10 Aug 2001, at 9:02, Randall Hofland wrote:
> 
> >     DynamicIP DHCP lease or not, I find all this "connectivity"
> > discussion most
> > interesting. While I wait impatiently for my own fast connection to
> > materialize
> > so I can set up my own operation, I seem to see that the service
> > providers are
> > working diligently to block the use of such connections for anything but
> > the most
> > mediocre and limited of home surfing options. Meanwhile, Verizon seems
> > to be
> > determined to prevent other options for home based e-commerce businesses
> > unless
> > they are overly costly and burdensome.
> >     As such, I am beginning to lean towards oppressive regulation of
> > telecommunications providers (baby bells and cable media companies in
> > particular)
> > so that they are absolutely forced to open up their backbone networks to
> > other
> > competitors rather than allowed to bleed to death those same competitors
> > as they
> > seem to have successfully done over the past few years.
> >     The open availability and reasonable pricing of internet options is
> > so
> > essential for small businesses in particular. We must ensure that a
> > megacorporate
> > mentality doesn't overwhelm that freedom. As such, I am curious just how
> > many out
> > there in BLU land have found their services for external users blocked
> > for some
> > reason by the broadband ISPs and the how's and why's it was done.
> > 
> > TIA
> > 
> > 

-
Subcription/unsubscription/info requests: send e-mail with
"subscribe", "unsubscribe", or "info" on the first line of the
message body to discuss-request at blu.org (Subject line is ignored).