From: jim bell <jimbell@pacifier.com>
Date: Sat, 18 May 1996 17:49:53 +0800
To: "E. ALLEN SMITH" <EALLENSMITH@ocelot.Rutgers.EDU>
Subject: Re: Past one terabit/second on fiber
Message-ID: <199605160153.SAA05576@pacifier.com>
MIME-Version: 1.0
Content-Type: text/plain


At 07:09 PM 5/15/96 EDT, E. ALLEN SMITH wrote:
>	One problem with the development of such high-end technologies is that
>they tend to increase economies of scale to the point where it's impractical to
>have anything but a monopoly or ogliopoly. As well as concerns about the degree
>of control such an organization may be able to exert in and of itself (acting
>like a government, in essence), there's also that such an organization is
>easier to pressure than a lot of small providers. Anyone have a suggested
>solution, or reasons that I shouldn't be so worried?

I think there are a number of reasons you shouldn't have to worry, at least 
about these technologies per se.  What these technologies do is dramatically 
lower the cost of supplying the service, both allowing the usual providers 
to cut prices and in fact forcing them (via the market) to do so.  Fiber, 
for example, lasts "forever," and effectively has an unlimited bandwith 
compared to most needs, which means that it has little cost other than the 
initial installation, amortized over time.

You might as well forget about this 1 Tb/s fiber, for example.  In order to 
justify such a thing, you need to have 1 Tb/s of information that you want 
to take from "here" to "there".  A city of 1 million people would have to 
have a 1 Mbit/sec/person data appetite to fill that fiber, and you'd have to 
have a huge data infrastructure just to collect all that data in the same 
place.  (Which would, in itself, require a mass of fibers of lower 
capacities.)   There would probably be no need for this, either, since the 
data would probably not all be coming from/going to the same location.  
Probably the only place where 1 Tb/s fiber will be needed in the next 20+ 
years is undersea links, and even that is questionable.

There's also the "all the eggs in one basket" problem.  Send your data on 
12, 80 Gb/s fibers and you have substantial redundancy.  Send it on one 
fiber and you're more subject to downtime due to individual component failure.

Secondly, fiber can be upgraded in an exceedingly economical, though 
granular way.  As I observed a few years ago, a given cable way initially 
had only a single cable installed though it had capacity for three.  And 
some of the fibers installed were probably not used immediately, due to lack 
of need.  Finally, the fiber was probably only driven initially at a 
comparatively low speed (maybe 400 megabits/s, for example) but could have 
been later upgraded to 2.5 Gb/s, or perhaps even higher.  Upgrading under 
those conditions is extremely cheap.

The main thing we need to worry about is allowing those people with the 
fiber to exert influence over us to an extent greater than the cost of 
supplying that data-transmission service would reasonably allow.  (The nosy 
landlord problem.)  Ideally, they'd provide the fiber, we'd pay for it at a 
reasonable rate, and they'd be satisfied.  The alternative, a busybody 
policy, is only practical to force on customers when the product being 
supplied is supply-limited.  Since fiber cost is dropping like a rock, and 
capacity will outstrip demand for the forseeable future, there is no reason 
we should be limited to deal with only one organization.

Jim Bell
jimbell@pacifier.com