In reply to a post by Andrue:

That old '1,024 is nearly 1,000' nonsense was never officially recognised.

It's still used in the computer industry.

Memory and disk, and hence data, are measured with K = 1024, M = 1048576 and so on, because it is easier and more efficient to build and address memory devices in powers of 2. After all they are really binary digital computers.

In reply to a post by XRaySpeX:

In reply to a post by Andrue:

That old '1,024 is nearly 1,000' nonsense was never officially recognised.

It's still used in the computer industry.

Memory and disk, and hence data, are measured with K = 1024, M = 1048576 and so on, because it is easier and more efficient to build and address memory devices in powers of 2. After all they are really binary digital computers.

Nah. In this day and age powers of two are more trouble than they are worth outside of embedded systems. It's far easier to work in powers of ten as it is in most things and on your typical computer it really makes no odds. The efficiency saving of having an array with 8,192 elements instead of 10,000 is miniscule and far outwheighed by the benefits of me being able to quickly calculate how much storage I need for six of them.

With modern hardware software developers are no longer tied to the old constraints. We can express ideas in terms that are better suited to our brains and SI is a lot easier than 2^x

SI units are what proper engineers use. It isn't 'cool' be to different in this case - just silly and inconvenient

That's a terrible example - 8192 is a *much* better figure to use, as this is memory and is *always* expressed using powers of two (just try buying a 8.192MB memory stick and see how far it gets you). Multiplying 8*6 in your head isn't exactly hard.

In reply to a post by Andrue:

(there's 8 bits in a byte)

As we are having the other debate I thought it a good time to point out (again) that this is not universally true.

Especially if someone's had a nybble at them

Cool or not, it's the way computers work internally, that I'm talking about. You are free to design and program your applications using decimals, and it would be stupid not to do so.

You talk as if software has only recently been designed to cater for decimals; far from it

In your example a memory of 10000 elements/words/bytes would need an Address Bus 14 wide (each capable of carrying +/- current or 0/1 or Yes/No). Consequently 6384 of the possible values would be wasted; i.e values 10000 to 16383 go to waste when they could be usefully employed at no extra cost.

BTW: One of the 1st computer systems I worked on in the '60s was BCD (binary coded decimal) based. It a memory of exactly 10000 storage chars that could hold the digits 0-9. We soon dropped such architecture LOL!

In reply to a post by TonyHoyle:

That's a terrible example - 8192 is a *much* better figure to use, as this is memory and is *always* expressed using powers of two (just try buying a 8.192MB memory stick and see how far it gets you). Multiplying 8*6 in your head isn't exactly hard.

It's harder than 10*6. I reckon everyone reading this post can immediately work that out in their heads - it's so simple that it's automtic. 8*6 OTOH requires thought for most people - especially those that are the product of the education system of the last couple of decades.

I'd also guess you've never been in a real-world software development environment. Sadly most of them are poorly run and/or not given enough time to do their job properly. Most developers either through laziness or lack of time will not bother to reach for a calculator. They will do their arithmetic in their heads and often take a wild guess at the answer. Working with powers of ten reduces the chances of mistakes.

For most applications the hardware requirements are irrelevant. Very few programs stress a computer sufficiently to require the developer to give a damn. 8192kb or 10,000kb makes no odds. You use as much storage as suits your purposes and it either fits or it doesn't. If it doesn't fit (which is pretty damn' rare) it will be by such a magnitude of difference that a couple of mega bytes is neither here nor there.

Personally I try to protect my team from crazy managers and do my best to give them the time and resources they need. Unfortunately I have to acknowledge the real world and compromise in some areas - perfect software is often last to market and flops as a result. If I found any of my team wasting their time trying to fit everything into power of two 'magic numbers' I'd have a quiet word them about it. The gains just aren't there. They'd get more benefit using the time to suf the web - at least they'd get a break from work that way.

In reply to a post by gomezz:

In reply to a post by Andrue:

(there's 8 bits in a byte)

As we are having the other debate I thought it a good time to point out (again) that this is not universally true.

That's true but I bet that 95% of developers, applications, users and computers are operating in a world where 8 bits equals a byte

In reply to a post by XRaySpeX:

In your example a memory of 10000 elements/words/bytes would need an Address Bus 14 wide (each capable of carrying +/- current or 0/1 or Yes/No). Consequently 6384 of the possible values would be wasted; i.e values 10000 to 16383 go to waste when they could be usefully employed at no extra cost.

Very few developers these days know what an address bus is. There's actually not that many that know what a byte is. I worry sometimes about the younger generation

..but then again it's probably what the old-school people have always felt as time moved on. The younger generation are simlply being saved from the hassle and baggage of the binary system under their fingers. They are free to expend energy in more productive areas. It's called progress. The work my generation did has laid a foundation on which the younger guys can build without having to reinvent the wheel. Talking to my Dad who spent time at ICL shows that his generation knew about things that I didn't.

Actually some of that is just sad. When I first heard about virtual memory I thought it was a great idea that someone at Intel or Microsoft had invented. Then I discovered that it was actually used in the 1970s. On the plus side - whilst I understand how Virtual Memory works I have never had to program paging registers on a CPU and I don't try and make my structures multiples of 4kb in size

Edited by Andrue (Tue 29-Sep-09 18:14:01)

In reply to a post by XRaySpeX:

BTW: One of the 1st computer systems I worked on in the '60s was BCD (binary coded decimal) based. It a memory of exactly 10000 storage chars that could hold the digits 0-9. We soon dropped such architecture LOL!

English Electric Leo III worked in BCD and had a hardware register which enabled calculations in £sd or imperial weights to be carried out easily. You just set that register to cause carry-over at the required places, and of course these were standard settings so nothing to think about re the required values.