User:Jtp15

From Wikipedia, the free encyclopedia

jtp15

My interests are in computer science and related areas such as networks and telephony. MS in computer science, BSEE. I started consulting in 1977 and have been putting food on the table developing software and systems ever since. I joined Mensa as a teen but got bored rather quickly, I don't enjoy discussions about how smart we all are, rather I enjoy discussions about what smart is.

I enjoy history, religion, philosophy, etymology and alternative energy, but mostly set theory and quantum states among my many eclectic studies. I tend towards existentialism but also lean towards the "Big Bang" being causal. One of my prescriptions says "take 1 pill as needed for mania" If this mean that I am a certified maniac, I am in denial, and repress it under a veil of wit.

As far as binary machines go; I am of the school of thought that we have reached the point where the circuits are small enough and spaced closely enough that external forces; electro-magnetic radiation, electro-mechanical irregularities and stray energy from static or power supply discrepancies combine to result in the dreaded "computer error". Add over clocking and you will get differences in one WORD (32 bit) rounding on symmetrically built computers.

In 1996, after building a two-tiered client-server based financial system, where the back-end performed the business logic and returned a consistent result set, my users showed me minor differences in the results on there work stations. This can not be; I thought.. I ran tests on machines I had ordered in a lot of twenty from a major micro-computer manufacturer that had the exact same build specifications. The lot even had sequential serial numbers, though of course not the components. After running many software tests, I ended up coding assembly language that ran directly on top of the BIOS (a mini-OS.). I put five of the work stations in a controlled environment to stabilize entropy. Of course there was still BIOS and the PROMs on the motherboard, but for practical purposes the conditions were far more stable than a typical PC in a cubicle.

I wrote a few simple rounding routines that each used hard coded constants 32 bit WORDs of rational and irrational infinites. Also a n depth stack that was n depth wide, where n was MAX(WORD). The result being a bi-cubic array where x=y=n. After y gigaflops the registers were read and I found that two of the machines were not able to be put in the same static state 23% of the time. In real world working conditions the accuracy was even less.

Thus, I will no longer blame the software for all errors. Though I never measured the frequency of incidence in more tightly controlled lab conditions, and did not narrow the results down to specific components or ROMs. My hypothesis is that even a HAL 9000 series is subject to a stray photon or two.

As you can see from this last blurb, I am cautious of absolutes. You might even call them one of my pet peeves.

So, except for a few minor idiosyncrasies, I would love to be of help with articles when I have the time or the need is great.

J. Ted Parker