[2434] in Commercialization & Privatization of the Internet
Billing systems
daemon@ATHENA.MIT.EDU (Jack Haverty)
Thu Feb 27 08:49:03 1992
Date: Thu, 27 Feb 92 05:44:25 PST
From: Jack Haverty <jhaverty@us.oracle.com>
To: geoff@fernwood.mpk.ca.us
Cc: jhaverty@us.oracle.com, drw@bourbaki.mit.edu, com-priv@psi.com
In-Reply-To: the terminal of Geoff Goodfellow's message of Wed, 26 Feb 92 19:59:54 PST <9202270359.AA07878@fernwood.mpk.ca.us>
Geoff,
Sure. I haven't thought about enumerating them, but here's a couple of top
ones:
- Saturation-behavior. The technology needs to function sensibly (in the users'
definition) when operating at the limit of some resource. My impression is that
the state-of-the-art is that whenever anything gets scarce, behavior similar to
"thrashing" in traditional time-sharing systems occurs, when little work gets
accomplished. The industry focus is on providing more resource - hence the
attention to T3 lines, giga-packet-per-second switches, etc. This also is one
of the drivers of demandwidth (I like it...). However, for many applications,
it is not appropriate to simply throw more resources at the task, but rather to
slow down the task (e.g., make the terabyte file transfer take longer), which
means the system operates at saturation. Currently, when this occurs, it's
common to see symptoms of thrashing - e.g., poor delay performance for
interactive users when the file transfer is in progress - and the schemes I've
seen so far (various kinds of priority queuing) provide only partial answers.
- Manageability - simply the ability to operate the overall system with specific
performance and cost requirements, using personnel who don't need to know all
the nitty-gritty technical details in order to fix problems, etc. How does
someone design/deploy/operate an internet (note I did not say The Internet) to
meet a well-defined set of performance specs (e.g., < 3 seconds to repaint a
screen 98% of the time) and cost specs (prove to me that the new design will
cost less to operate than the old SNA network)
- Parochialism -- there's lots of other networking systems in use (Netware,
Vines, XNS, X.25, Decnet, SNA/APPC/APPN, Appletalk, ISO, etc., etc., etc.) and
in a typical corporate environment you'll find several if not all of them.
TCP/IP is another player on the field. If you talk with any of the proponents
of any one of these, in general you'll find a common perspective - that their
technology is the primary one which will be used, and various schemes will be
offered to support other religions in limited ways (e.g., encapsulating
packets). I think this is partly the cause of the two topics above.
Enough for now....
Jack
