Posts by Tony Chesser

Metric versus imperial: Reg readers weigh in

As a yank...

Yes, I know, "Yank" is a Brit term for someone in the US. This is a Brit site. When in Rome ...

I have no problem with going SI. Indeed, physics and chemistry courses, offered at the college level, are all SI. We don't talk about ounces of this dissolved into gallons of that. We don't deal with 32 feet / sec^2 for gravitational acceleration. We deal with grams, liters and 9.8 m / sec^2. Trying to apply physics in daily life tends to be difficult, as we have no frame of reference for meters / sec on velocity, being used to miles / hour.

Frequently, I've found that a set of metric (SI) wrenches is workable for both standard and metric nuts, bolts, etc. I'm much more likely to find a metric wrench which is "close enough," when dealing with standard (Imperial) bolts, than I am to find a standard wrench which is "close enough" for a metric bolt. This is, of course, because the divisions are much smaller in metric (1 millimeter) than in standard (1/16 of an inch).

There's only one problem with SI units: how do you get an even 1/3 of anything? A lot of imperial units are based around multiples of 12, which is evenly divisible by 2, 3 and 4. 12 inches in a foot. 60 seconds in a minute (5 * 12) or minutes in an hour. 24 hours in a day (2 * 12). 5280 feet in a mile (440 * 12; you get divisions by 5, 8, 10 and 16 in the bargain).

Working with computers, as I do, we run into a completely different problem. 1/10 is NOT exactly described by an IEEE-754/854 standard floating-point number. Ergo, any computer is going to either use some ancient BCD format to describe it (and describe it exactly) or it will have to store an approximation of it (more commonly the case). In that respect, Imperial is no better than SI because there's no exact match for 1/3 (or 1/12), either.

Wondering if we need a numbering system built around divisions of 30, so that we can evenly divide into 2, 3 and 5.

Note: I work with IBM proprietary hardware, quite often. The CPUs in those things support BCD formatted numeric data and some of the languages, such as RPG and COBOL, actually prefer that format for numeric data. Ergo, those support SI, even though that's considered a "legacy" numbering format.

IBM euthanizes Power6 and Power6+ systems

Don't have source code?

IIRC, one of the "features" of OS/400, i5/OS. IBM i or whatever they're calling it this month, is that object code is stored in two forms: a VM-type object code and native platform object code. Consequently, when you move from Power6 to Power7, and you migrate your partition, it throws out the native platform object code and rebuilds it from the VM-type object code. This is so that you don't have to port the source code to the new hardware. If you had to actually port the software to new hardware, every time you did a hardware upgrade, you might be tempted to upgrade to a different hardware vendor while you were at it :-) Clearly, IBM would rather you didn't do that. Hence the reason for this feature.

So, why is it that you need the source code?