This person can no longer be contacted through the School of Engineering and Computer Science at Victoria University of Wellington
Computer geek. Soldering by 7. Started programming at age 10 and burned out by age 14. Locked out a school Novell NetWare server (even to the superuser) at 13. By 17, was running 486 motherboards on sheets of newspaper in the school library, and experimenting with bridging jumpers with screwdrivers to see what would happen when the connection was severed while the machine was running (changing bus wait states causes a stack dump in DOS, by the way).
I have six years of professional systems administration experience, primarily in Linux/UNIX open-source systems.
My professional experience has given me a perspective on real-world systems management. Having a hatred of needless repetition, I can't stand administrative tasks which could be automated, and log filtering is one of these. The vast majority of log filtering is done with long lists of regular expressions which have to be maintained so that they match the particular versions of software which are installed at any given time, and upgrades (whether for features or security purposes) often entail a bothersome process of updating the log filters. Other methods such as Baysian filters or data mining also have to be "re-educated" in these circumstances. Millions of administrators around the world have to perform the same tasks in parallel every time they change their systems.
Moreover, the messages which are presented to an administrator on a daily basis often have no salience value. Whitelists and blacklists don't care about salience. I aim to offer a scheme whereby this can be improved and repetition kept to a minimum. A multi-dimensional, human-and-machine readable scale is currently the most feasible alternative I can envisage.