Subject: Re: The future of ISPs From: Rex Ballard Date: Mon, 26 Jun 1995 15:34:05 -0400 (EDT)
How the Web Was Won
Subject: Re: The future of ISPs From: Rex Ballard Date: Mon, 26 Jun 1995 15:34:05 -0400 (EDT)
In-Reply-To: <93950611202039/0006058685NA3EM@MCIMAIL.COM>
Message-ID: 
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII



On Sun, 11 Jun 1995, Curt A. Monash wrote:

> Could somebody please review the extent to which Net economics are still
> "artificial"?  

Internet economics break down to two factors.  First, the ability to 
utilize each link with extreme efficiency.  Second, the ability to 
minimize the administration overhead through automated systems (routers).

In terms of the actual costs to use the "wire", there are a certain 
number of bits you can put on a wire for a given second.  For copper, the
limit is about 1 million bits/second for a 1 kilometer span.  For 
fiber-optics, the speed is closer to 5 billion bits/second.

Analogue audio, especially modems, make incredibly inefficient use of the
available wire.  The line must be sampled 8000 times/second (64,000 bits)
whether you are downloading a video clip or in the bathroom.  The analogue
signals are combined with other streams.  These streams are time-critical
and must be delivered with a minimum of delay.  The cost for a dial-up line
from New York to California is slightly under $10/hour per circuit.

With internet protocol, you only use the bits you need.  If you have 2000
bytes of text and 4k for a 64 bit x 64 bit icon, you will need about 56 Kbits
for a basic web page.  Now you will spend 30 seconds looking at it.  As a 
result, you only need 120 Kbits/minute instead of 64 Kbits/second.  To use
an hours worth of equivalent voice bandwidth, you would need to download
almost 500 web pages.

With IRC, you can use compression techniques to reduce your voice bandwith 
from 64 kbits/sec to under 8 kbits/sec.

> If long-distance costs are essentially "free", then somebody's mooching on
> somebody, and that will unravel as the numbers become too big to ignore.

Actually, there is an averaging effect.  For $20/month, I can receive
about 100 megabytes of what ever I want.  In fact, because I'm not 
forcing the provider to bill me only for my share, I can download about
500 Meg for $20 and the ISP and the Long distance companies will still make
out like bandits.

If I have direct access to a T1 circuit, I have eliminated the need for 
an intermediate host and modem banks.  This now makes it possible to 
cost-effectively deliver at under $2/gigabyte.  The internet backbones
carry over 5 gigabits/second on each of their 25 strands.

> But I'm not at all current on the details, and I think they're relevant to
> the AOL vs. Netcom vs. local ISP discussion.

The main difference between AOL, Netcom, and pure ISP access is the level 
of support services required.  AOL provides dialups, disks, access to 
remote servers, and help personnel 24 hours/day 7 days/week.  My local
ISP might provide a "almost local" dial-up, a sparcstation shared by 500 
users, and e-mail responses to questions within 5 business days.  For AOL,
I pay $3/hour for internet.  From Digex, I pay $30/month and never use
all of my time/data/disk quotas.

Of course, when I run my PC in Linux mode, I become the server and gateway.
At that point I don't have to keep anything but my "outbound" (web pages)
on their server.

At work, I share a T1 with about 500 other people.  Reponse times get a 
bit slow when 50 of us are on at the same time.  The carrier gives us a 
very good rate, and we never use our full quota either.

> Curt Monash

	Rex Ballard
	Standard & Poor's/McGraw-Hill
	Opinions expressed do not necessarily reflect
	the Management of the McGraw-Hill Companies.




From rballard@cnj.digex.net Mon Jun 26 16:03:09 1995
Status: O
X-Status: