From: jim bell <jimbell@pacifier.com>
Date: Tue, 11 Feb 1997 06:12:18 -0800 (PST)
Subject: Re: ISPs vs Bells - Email FCC
Message-ID: <199702111412.GAA19446@toad.com>
MIME-Version: 1.0
Content-Type: text/plain


At 11:26 AM 2/7/97 -0500, Derrick Storren wrote:
>By Michelle V. Rafter 
>LOS ANGELES - Regional phone companies and Internet service providers
>are waging a war of words over Internet traffic on the nation's local
>telephone network and who should pay for upgrades as the online boom
>continues. 
[snip]
>In one corner, Pacific Bell, Bell Atlantic and other regional phone
>carriers say the growth of Internet traffic is pushing local telephone
>networks to the breaking point. 
>
>Pacific Bell, for example, says Internet surfers use its phone lines an
>average of 45 minutes a day -- more than twice the amount the network
>was built to handle. Heavy Internet use in Silicon Valley led to brief
>service outages in that area earlier this year, the company said. 

This is an "excellent" example of lying with statistics.  Contrary to the 
implication above, telephone switches aren't designed to handle a certain 
amount of telephone traffic per day; Rather, they are designed to handle a 
certain peak amount of traffic.  This, and the knowledge about the typical 
usage patterns that communities generally see in their telephones allows a 
statistician to estimate how much traffic that switch will actually see per 
day, in the real world.

However, change the pattern of usage, perhaps by adding usage to 
previously-underused time periods, and you could dramatically increase the 
daily traffic statistics for a given telephone switch.  If, say, 2% of the 
population were to suddenly decide to make 6-hour phone calls daily between 
12 midnight and 6am, a time of very low usage, you might increase the 
average daily usage by 50%, but with absolutely no increase in the peak 
usage during the daytime hours nor need for new switches.

While that's a fanciful example, a similar effect occurs, I think.  I've 
seen a set of graphs showing the typical usage level for the Teleport ISP, 
and it appears that while usage reaches a level of about 70% between about 8 
am and 4 PM, it continues to increase after 4pm, solidly peaking between 8 
pm and 11 pm at about 98% usage.  This is long past the time that most 
humans make voice phone calls.  The implication is that easily 3/5 of 
Teleport's traffic occurs after 4 pm and 6 am next morning, a time frame 
which is definitely post-peak hours.   (Traditionally, pre-Internet, 
telephone usages peaks at about 11 am and 3 pm.)

Since modern telephone switches don't wear out, unused call capacity is 
simply wasted.  It makes no sense to charge people more for services which 
cost the supplier no more to provide.


>If, for example, Internet providers passed through an access fee of 1
>cent a minute, a subscriber spending 10 hours online a month would pay
>an extra $6 -- hardly a deterrent, said David Goodtree, an analyst with
>Forrester Research in Cambridge, Mass. 

Even that is unacceptable.  If an average ISP's phone line was busy 50% of 
the time, with "only" an extra charge of 1 cent per minute, that would be an 
average charge of $7.20 per day, or $216 per month, as compared to a typical 
business line which might cost, say, $40 per month.  


Jim Bell
jimbell@pacifier.com