[1553] in Commercialization & Privatization of the Internet
Re: Growth of the Internet ...
daemon@ATHENA.MIT.EDU (Kent W. England)
Mon Nov 4 18:25:58 1991
From: "Kent W. England" <kwe2@BBN.COM>
To: dennis@mrbill.canet.ca
Cc: com-priv@uu.psi.com, dotytr@nscultrix2.network.com
In-Reply-To: <91Nov4.150346gmt.105469@MrBill.CAnet.CA>
Date: Mon, 4 Nov 91 18:18:55 EDT
>The problem which occurs when speculating on the size of the Internet
>is that there is no real way to measure the things people are really
>interested in (number of users, number of hosts, that kind of thing).
How about a measure like "correspondence distance"? Ie, for a given
prolific internet correspondent, how far is every internet-connected
person from having directly corresponded with said author? If someone
has directly sent a private message to or received one from Prolific
Correspondent their distance is one. If someone has corresponded with
someone else who has directly corresponded with Prolific Correspondent
their distance is two, and so on. This is the "Erdos size" of the
Internet, as described in Carl Malamud's book "Stacks" which was
available at InterOp this year.
Carl tells the story that Mike Schwartz from Colorado did the same
analysis on the From and To lines on mail messages and found a diameter
for the Internet of about 12. It might be closer to 6 or 8 for a more
complete sampling, according to the story.
Is this an interesting measure of the size of the Internet? Yes, but is
it useful? I don't know. But if someone tracked the size of the
Internet growing past the small world paradigm, that would be a
significant measure of truly gargantuan proportions. And at least the
Erdos number can be approximately measured, unlike some other measures
of the size of the Internet (ie, connected hosts or users). Of course,
measurability is one reason we all like the connected set of networks as
a measure. But I think Erdos size is, well, cuter. :-)
--Kent
