[2202] in Commercialization & Privatization of the Internet
Low end users
daemon@ATHENA.MIT.EDU (Barry Shein)
Thu Jan 30 17:32:11 1992
Date: Thu, 30 Jan 92 17:30:53 -0500
From: bzs@world.std.com (Barry Shein)
To: com-priv@psi.com
In-Reply-To: drw@BOURBAKI.MIT.EDU's message of Thu, 30 Jan 92 16:51:49 EST <9201302151.AA04578@lagrange>
Re: advantages of centralized systems
However, the disadvantages of centralized systems are:
1. Lowest common denominator support, resistance to
customization of environment (sorry, we only support
the XYZ spreadsheet, sorry, we don't have many customers
interested in XYZ so we have no packages nor plan to
in the near future, etc.)
2. Response time - You can decide how the cycles on your
desk are best to be used for your own purposes.
3. Resource fascism - resources are shared so unless there
are infinite resources, centrally, some central administration
must make resource policies. These tend towards artificial
and real pecking orders (usually artificial in my experience,
with center staff giving themselves huge amounts of resources
on demand while trying to suppress usage by those involved
in the actual mission of the organization.)
A good example of this: Right now a 1GB disk is under $2k.
Go survey centralized academic systems and ask what it would
take to get 1GB of disk, for your personal use, on their
centralized system.
4. Similarly, us/them mentality - Central administrators come
to view the user base as the enemy, related to #3. This is
really the source of many other problems and can be an
overwhelming problem, corrupting every decision made.
5. Where applicable, centralized systems can become powerfully
manipulative tools (I heard of one case where a faculty member
was suing the university (a tenure issue), so w/o warning
they just shut off all her accounts and removed her files,
she lost access to a manuscript she was writing under deadline,
they could do it, so they did, who do you appeal to?)
Some of these problems can be resolved by usage pricing, that is,
making users of centralized services *customers* rather than "users".
A lot of the imagined efficiency of support is just that, imaginary.
Centralized systems I have been involved with had as their primary
mission (it seemed) to just grow and grow without obvious bound. And
to reduce services simultaneously. They became political power
centers.
Typically, this is done by denying responsibility for new, distributed
facilities which were actually replacing many of the centralized
facilities.
For example, printer services. The central printers ceased to be used
as much (often because they didn't support the software the
distributed systems used) but the reduction in central paper (etc)
expenses were viewed merely as a boon to the center.
I believe that it is the above sort of abuse, rather than other more
positive perceptions of distributed systems (other than inasmuch as it
helps elminate the effect of the abuse), that people ran to
distributed systems when they had the opportunity.
However, proper political structuring (again, making people customers
and the success of the centralized organization critically tied to
their satisfaction) can solve most of these problems.
But in academic settings this is rarely if ever the case (outside of a
few major grant holders), and no doubt the reason for the success of
even half-assed distributed systems.
-Barry Shein
Software Tool & Die | bzs@world.std.com | uunet!world!bzs
Purveyors to the Trade | Voice: 617-739-0202 | Login: 617-739-WRLD
