[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Troubleshooting Wizards for Linux (fwd)
Anyone on Linux-IL looking for an idea for a novel product for a startup
business?
---------- Forwarded message ----------
Date: Tue, 10 Jun 1997 21:49:14 -0600
From: Kevin S. Van Horn <kevin.s.vanhorn@iname.com>
To: cons-forum@redhat.com
Subject: Troubleshooting Wizards for Linux
One way of improving support for Linux would be to provide a utility to
diagnose and suggest fixes for common problems. Microsoft has done this
for Windows '95, with their various troubleshooting wizards. These ask
a series of questions of the user, and based on the responses they ask
more questions or suggest possible fixes. It's a nice facility for an
otherwise abominable OS :-)
The Microsoft troubleshooting wizards are based on Bayesian belief
networks (BBNs). BBNs can be thought of as an expert system technology;
their advantage over older approaches is that they have a theoretically
sound basis (probability theory) for encoding "fuzzy" or uncertain
knowledge and reasoning with partial or uncertain information.
A BBN consists of three things:
(1) a directed, acyclic graph whose nodes are variables or attributes
of the system being modeled;
(2) a finite set of possible values for each node; and
(3) for each node, a conditional probability function that gives, for
every possible combination of values of the node's parents,
the probability of each possible value of the node.
A BBN for Linux troubleshooting would have nodes representing various
aspects of the system state, and others representing various problem
symptoms. The arcs and conditional probability functions can be thought
of as representing causal relationships.
I know enough about BBNs to create a troubleshooting utility operating
on a Linux knowledge base. What I lack is sufficient Linux expertise to
create the knowledge base by myself. (I'm a long-time Unix user, but
was never a sysadmin, and only recently started using Linux.) To create
the knowledge base I would need the opportunity to interview some
experienced Linux troubleshooters. It would also be very useful to have
a database of case histories: records of particular problems solved,
listing the problem symptoms noted and what was eventually determined to
be the problem cause (or at least an action that fixed the problem).
Would RedHat -- or anyone else on this list -- be interested in working
together on such a product?
----------------------------------------------------------------------
Kevin S. Van Horn
kevin.s.vanhorn@iname.com
--
To unsubscribe: mail cons-forum-request@redhat.com with
"unsubscribe" as the Subject.