Subject: RE: Web Count From: R Ballard Date: Mon, 10 Apr 1995 20:23:55 -0400 (EDT)
How the Web Was Won
Subject: RE: Web Count From: R Ballard Date: Mon, 10 Apr 1995 20:23:55 -0400 (EDT)
In-Reply-To: <199504100021.UAA03048@gold.interlog.com>
Message-ID: 
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII



On Sun, 9 Apr 1995, Jay Linden wrote:

> At 12:30 PM 4/6/95 -0400, etta palm wrote:
> 
> >On Wed, 5 Apr 1995, R Ballard wrote:
> >> On Wed, 29 Mar 1995 ADAMX10415@aol.com wrote:
> >> There are about 36 million internet users, most of whom have web 
> >> capability of some form (Mosaic, Webster, Cello, NetScape...).
I picked the following information off of InterNIC, the agency that 
administers internet addresses.

Class A Adresses  (Potential 16 million hosts each) 116
Class B Addresses (Potential 65 thousand hosts each) 9992.
Class C Addresses (Potential 256 hosts each)	559807

Assuming 1 host per subnet, I still come up with over 10 million hosts.
Class C Subnets are carefully watched and consolidated.  Many of these 
may be protected through fire-walls.  If you estimate capacity of those 
addresses currently assigned, there is the potential for 2.5 billion hosts.
Somewhere in between is the real number  :-).

> Most of the "best-guesses" I've discussed involve a somewhat lower 
> number; of course, most of them are a few months old too and it's entirely
> feasible that the others were fairly accurate then and that 36M is fairly
> accurate now.

Since the internet has no central registry of users, and most authorizing 
agents are reluctant to give out detailed information about their users, 
it is increasingly difficult to know how much of what people are using.  
Even statistics such as tcp-connects through the NSF backbone and k-bytes 
of traffic recieved are less reliable since there are now about 20 
international and transcontinental supernets.  Just about every common 
carrier capable of passing an IP frame is carrying IP packets, sometimes 
without even knowing it.

At some point, when the net was at about 20 million active accounts, the 
common carriers started splitting up the internet market.  Since IP 
frames can be spread across multiple carriers by the load balancing 
routers, there may be no accurate method of authoritatively pegging the net.

> My caution is that the net is simply very difficult to count, and that
> nobody has tried to count it scientifically, having and using the 
> necessary resources to do so.

Short of including it in an international census (remember, the internet 
includes almost every country in the world).  It is highly unlikely we 
could ever get an accurate count without forcing all carriers to put 
sophisticated monitoring equipment on the back-bones, and forcing them to 
give out highly confidential competitive edge type information.

> To this juncture, it has been difficult enough to even establish a
> definition of what groups of people are actually "on the net" -- e-mail
> accounts?  Some extra (ftp, telnet, gopher) capabilities?  Shell account
> with lynx?  Slip/PPP account?  Hosts only?  How about the Fidonetters?

There are actually about 6 classes.
	E-Mail Only
	E-Mail, News, FTP (PSI Link)
	Shell - commends from a UNIX or VMS host.
	TIA/FireWall (Protected clients calling out through what appears 
			to be a single host)
	SLIP/PPP Dial-up (Hosts which are only on-line when user wants to 
		be connected) often, 100 users will share a pool of 20 hosts.
	PPP/T1+ (Hosts which are up 24 hours/day 7 days/week)

There are economic factors.  When collecting $30/user, you will need at 
least 30 users to cover the costs of 2 T1 spans.  To recover capital 
costs and break-even within 12 months, you need 60 users/host.

> There are a couple of major surveys being undertaken presently; of
> course, their results are in danger of being obsolete before they're
> published.
We've been trying to keep statistics on the internet for 15 years.  It 
would be a real challenge to see how an authoratative survey could be 
done on this network of networks.

> I have a tremendous respect for Rex, Adam and others on this list who
> have been around longer than I have, who have a better knowledge than I
> of the net's history and technology, who have worked very hard to make
> the net a place where I love hanging out and doing business.
The internet is an unruly 


> I just tend to be cautious of any numbers which are not, and cannot be,
> derived by anything other than an actual head-count.  Whether they're
> accurate or not, they're still just educated guesses.  When it comes to
> net and web population, there's a huge diversity among the educated
> guesses of educated people.

We have statistics on how many news-groups there are (publicly accessible),
how many postings are made to those newsgroups, and how many different 
e-mail addresses have been used in the last 12 months.  We know how many 
downloads of Netscape occurred since it's release.  We know how many
people hit popular web home-pages.

How many people walked through Manhatten between Noon and 6 PM Sunday?
Did you count all the kids who walked under the turnstile?  Did you count 
the people who only changed trains?  How many people where there in each 
care that went through the tunnel?  At 2:00 AM on Sunday night, Manhatten 
looks like a ghost town by the WFC.  At 9:00 AM Monday, there are so many 
people you can't even count them.  They pay about $2.00/person to enter, 
and pay about $2.50/person on transit once they are inside.  This is an 
island with toll-boths at every entrance.

How do you count the number of "commuters" on the internet, where there 
are no toll booths or accounting systems, only IP Packets flowing 
anonymously through hundreds of high-speed links.  It's more like trying 
to figure out how many people walked through Denver Colorado (a town 
surrounded by land, fed by 6 free-access interstates which are routinely 
used for cross-town commuting.  I wasn't even counted in the Census, 
neither was anyone else living in "rent by the week" housing in Denver 
that month.


> Jay Linden                                          Phone: (416) 510-8948
> Toronto, Canada                                 Fax/Modem: (416) 510-8949

	Rex Ballard
	N.Y.N.Y.


From rballard@cnj.digex.net Mon Apr 10 20:29:20 1995
Status: O
X-Status: