[2107] in Commercialization & Privatization of the Internet
re: Trickle Down? How About Buggle Up?
daemon@ATHENA.MIT.EDU (Craig Partridge)
Mon Jan 27 12:09:55 1992
To: Jack Haverty <jhaverty@us.oracle.com>
Cc: com-priv@psi.com
Reply-To: Craig Partridge <craig@aland.bbn.com>
From: Craig Partridge <craig@aland.bbn.com>
Date: Mon, 27 Jan 92 09:05:54 -0800
> "...It is possible to do dial-up IP just as easy as it is to do uucp or FIDO
> these days...."
>
> This is true, except that you can't exchange mail or files between TWO dial-up
> users without some effort to make sure they're both online.
Jack:
Not necessarily true. Some dial-up IP implementations require that
users manually bring up their lines. Others will actually arrange to make
the connections without any intervention (if a packet arrives and the link
is down, a phone call is made). Indeed, the first dial-up SLIP package
done by CSNET did connections on-demand.
Note that to make automatic dialing work, the dial-up hosts have to
have permanently assigned addresses (if a packet arrives for X.Y.W.Z, I
need to be able to deterministically translate X.Y.W.Z into the right
modem number to call). Some network providers believe for a variety of
minor logistical reasons that it is more efficient to share a small pool
of IP addresses among multiple users and assign a new IP address to a host
every time it is connected. [A nice little com-priv example of a case
where the better service exists, but for commercial reasons, often isn't
offered].
Craig
PS: One other observation -- some TCPs still have a built in 30 second
timeout for connection establishment. That doesn't quite work for a
dial-up world. Making a modem connection typically takes between 15
and 30 seconds. So my automatic mail delivery scenario between two
dial-up hosts doesn't quite fly unless one's TCP is up to snuff. (Another
reason why providers don't offer the better service).
