>Jim;
>
>Sorry to jump in but I could not help it.
>
>Personally, I think of the resolution of software errors to be similar to an asymptote. You may attempt to approach zero but you will never get there!
>
>Another thought I have on this subject is that we do not know all the existing errors but we can say that we test to 80% of known errors. The remaining 20% take a great deal of effort to resolve and it may not be economically feasible to consider. If only the most serious errors are addressed and those that affect the greatest audience then a good product has been delivered. In the end it is a management decision to resolve those errors it deems critical and important.
>
>Here is an analogy: Software is like hardware in that neither will ever be perfect. In the hardware world we allow tolerances and uncertainties. Tolerances exist because of the imperfection of humans. Plus or minus one micron as an example, is a tolerance (a very slight one at that!). Gears have a tolerance and will work within specified limits. The teeth will never have a perfect mesh or dimensions but yet gears can work. Uncertainties exist because of human inability to measure with total accuracy.

When working with large numbers of atoms, it becomes impossible to control each individual atom. For instance, the silicon used for transistors, etc., has to have an amount of controlled impurities less than one millionth. Since this can't be controlled exactly, commercial transistors of the same type may have very different characteristics (the "beta" number, which indicates the amount of amplification under certain conditions, may vary from 100 to 900 in one particular model). Designers have to put up with these variations, and seek to offset it in the complete circuits.

On the other hand, when working on a smaller scale (individual atoms), it turns out that nature itself is, in a way, "imprecise". By Heisenberg's Uncertainty Principle, you can work with probabilities, but never have absolute certainty of where an individual electron will go, for instance.

>I am taking my examples from the physical world of Metrology, the science of measurement. This was my previous career.

I just wonder, how in the World did you change to software development?

(Similarly, you might wonder how I ended up teaching computer networks, when my previous job was in computer programming! I had a job offer, I explained frankly what I knew and what I didn't, but my current boss thought that whatever I needed to learn, I would learn in the instructor training - they needed an instructor urgently, and I needed a job!)