Kilobyte

[Moshly]

This is so dumb and has done my head in ever since I heard of it.

In decimal 1K = 1000 (nice round number) 1000 in binary = 11 1110 1000 <- not very rounded.
In binary 1KB = 100 0000 0000 (nice round number but long) or in decimal 1024 (not so nice)
In binary 1MB = 1 0000 0000 0000 0000 0000 (nice round number but very long) or in decimal 1048576 (not nice but a lot shorter)

Its a bit like the difference between  1/3 and 0.333333333333333333333333->
One is nice & neat, the other stretches off to infinity.

[hamster_nz]

This isn't stackexchange, so let's restate the context...

Context is everything when arguing that "kB" strictly means 1000 bytes and not 1024 bytes... which was my point in mentioning 'k' was also used in the SI system for the Boltzmann constant. The same symbol can mean different things in different contexts, but you have to know the context you are using it in.

However, I do strongly agree that KHz is wrong, but close enough for jazz and nothing to get worked up over. 

[magic]

KHz is obviously the SI unit of rate of change of temperature.

[madires]

How interesting that you seldom see these new prefixes in literature. Haven't seen one in anything consumer, prosumer, nor professional. Not in device spec sheets. Not in technical ads for any component nor system. Haven't heard them come up in any discussion, in person nor online, except this thread. If they're catching on, they're hiding it well!

I've written a software package which outputs file sizes as part of its job. Six years ago I added configuration switches to allow each user to select his preferred prefixes. The default setting is classic binary prefixes with the multiplier of 1024. First alternative is SI prefixes with a multiplier of 1000, and the second is IEC prefixes (Ki/Mi/...) with a multiplier of 1024. This isn't anything brand new, still not many are aware of it.

[tggzzz]

This isn't stackexchange, so let's restate the context...

Context is everything when arguing that "kB" strictly means 1000 bytes and not 1024 bytes... which was my point in mentioning 'k' was also used in the SI system for the Boltzmann constant. The same symbol can mean different things in different contexts, but you have to know the context you are using it in.

However, I do strongly agree that KHz is wrong, but close enough for jazz and nothing to get worked up over. 

Just so.

But in the context of this thread, it is worth getting that particular LART out

[rsjsouza]

It takes time but it will be used.
Now wait till the americans understand they need to put the little i inbetween instead of reverting back to their KB and call that a KibiByte 

At a certain point in time I have seem capital K as "kibi", thus this may only attest to the ruling confusion in the wild.

The first computer I used had (and has - there is at least one still operating) 39-bit words.

That's a very strange number of bits. Systems like the PDP-10 and many IBM mainframes used 36 bit words since it divided evenly into groups of 6 bit characters. Other systems used 12 or 18 bit words for similar reasons.

But what on Earth had 39 bits?

Elliott 803/903 and similar. They were commercially important, and CAR (Tony) Hoare (I'm sure you know that name) developed a highly influential Algol60 compiler for them.

The world's first computer used for commercial operations had 35 bit words. (LEO 1, used from 1951 for stock control in teashops - seriously).

CDC 6600 had 60 bit words, ICL1900 had 24 bit words.

No doubt there are many others.

The F28x family of DSPs from TI have 16-bit bytes. A word would be 32-bits.

In the x86 world, it used to be customary to mention that 16-bit was a word, with the cousins Double word (32-bit) and Quad word (64-bit). I am not sure if that is still the case (I haven't revisited the x86-64 assembly).

The microchip MCUs that I worked with had 14-bit bytes on the program memory bus.

[tggzzz]

It takes time but it will be used.
Now wait till the americans understand they need to put the little i inbetween instead of reverting back to their KB and call that a KibiByte 

At a certain point in time I have seem capital K as "kibi", thus this may only attest to the ruling confusion in the wild.

I hadn't seen "kibibyte" before this thread.

It reminds me of a meme from a couple of decades on usenet... people objected to the use of "he" and "she", invented "hir" as a gender-neutral word and tried to insist everybody used it. Clearly it had escaped them that there was an existing alternative that worked perfectly well: "their".

The first computer I used had (and has - there is at least one still operating) 39-bit words.

That's a very strange number of bits. Systems like the PDP-10 and many IBM mainframes used 36 bit words since it divided evenly into groups of 6 bit characters. Other systems used 12 or 18 bit words for similar reasons.

But what on Earth had 39 bits?

Elliott 803/903 and similar. They were commercially important, and CAR (Tony) Hoare (I'm sure you know that name) developed a highly influential Algol60 compiler for them.

The world's first computer used for commercial operations had 35 bit words. (LEO 1, used from 1951 for stock control in teashops - seriously).

CDC 6600 had 60 bit words, ICL1900 had 24 bit words.

No doubt there are many others.

The F28x family of DSPs from TI have 16-bit bytes. A word would be 32-bits.

In the x86 world, it used to be customary to mention that 16-bit was a word, with the cousins Double word (32-bit) and Quad word (64-bit). I am not sure if that is still the case (I haven't revisited the x86-64 assembly).

The microchip MCUs that I worked with had 14-bit bytes on the program memory bus.

I thought of seeing what some of the harvard architecture MCUs did, and decided my examples were sufficiently strange

[Canis Dirus Leidy]

Whole thread in a single question of RU.OS.CMP FAQ:

Q50: And what does a byte look like?
A50: Ever looking through a punch card?

[splin]

Whole thread in a single question of RU.OS.CMP FAQ:

Q50: And what does a byte look like?
A50: Ever looking through a punch card?

The answer looks like it could be amusing, profound or perhaps sarcastic but only speaking English I've no idea what it means.  Can anybody translate?

[MT]

It takes time but it will be used.
Now wait till the americans understand they need to put the little i inbetween instead of reverting back to their KB and call that a KibiByte 

At a certain point in time I have seem capital K as "kibi", thus this may only attest to the ruling confusion in the wild.

I hadn't seen "kibibyte" before this thread.

It reminds me of a meme from a couple of decades on usenet... people objected to the use of "he" and "she", invented "hir" as a gender-neutral word and tried to insist everybody used it. Clearly it had escaped them that there was an existing alternative that worked perfectly well: "their".

Kibi was new for me as well, in Swedden Hir or as they say here "hen" been implemented for some years now particularly and frequently used by fake news and their so called journalists, symboled like this, giving the picture that
if a she and hi copulate they produce a hen, i.e  a probable soyboy gender bended trans bi pan sexual creature of sort!

[Nusa]

Whole thread in a single question of RU.OS.CMP FAQ:

Q50: And what does a byte look like?
A50: Ever looking through a punch card?

The answer looks like it could be amusing, profound or perhaps sarcastic but only speaking English I've no idea what it means.  Can anybody translate?

It has to do with the historical definition of the byte, coined in the late 1950's, which was originally defined as the number of bits used to represent a character. On IBM punch cards, which are way older than that definition, the most common version of which had 12 rows x 80 columns, each column represented a character. Or a 12-bit byte. You can look up Hollerith code if you want to know more about that. The code was designed not to put too many holes in the card, since the card would lose physical integrity and jam card-readers if you did so.

Many of the early computers did not use 8-bit bytes. 6 was actually quite common for a while, although there were other sizes. The IBM 704 had a 36-bit word and 6-bit bytes. The PDP-8 had a 12-bit word and 6-bit bytes. The Cyber 6600 had a 60 bit word and 6-bit bytes. And if you ever wondered why languages have provision for octal as well as hexadecimal, it's because 2 octal digits are a convenient human-readable representation of a 6-bit byte.

It wasn't until the processors of the late 1970's became widespread that the general use of the word became specifically 8 bits.

[fourfathom]

It wasn't until the processors of the late 1970's became widespread that the general use of the word became specifically 8 bits.

So does *anybody* these days use "byte" to mean anything other than 8 bits?  Personally, I'm not planning to travel back in time, so I'm comfortable with the 8-bit byte.  And with K = 1024 when used in context.  Context is critical.

[GeorgeOfTheJungle]

Gimme a few more bytes, honey.

[ggchab]

It wasn't until the processors of the late 1970's became widespread that the general use of the word became specifically 8 bits.

So does *anybody* these days use "byte" to mean anything other than 8 bits?  Personally, I'm not planning to travel back in time, so I'm comfortable with the 8-bit byte.  And with K = 1024 when used in context.  Context is critical.

The French translation of the term "byte" is "octet" which clearly contains "8" 

[magic]

And the English translation of "octet" is "octet" too.
That just means that the French have no concept of byte at all.

[ggchab]

I did not know the word "octet" was also used in English !
But you're right: the exact French translation of "byte" is "multiplet" and "octet" is a byte of 8 bit.

[Jan Audio]

I always wonder why it is 8 bit everywhere, i dont complain just wondering.
Compatibility issues ?

[magic]

You will see a lot of "octets" when you start reading computer networking standards. Many of them date back to dinosaur era; some well-defined and unambiguous unit  was required to specify how data are to be transported over network and "byte" certainly wasn't that

[ggchab]

I always wonder why it is 8 bit everywhere, i dont complain just wondering.
Compatibility issues ?

Maybe because some systems used BCD (binary coded decimal) in the past ?I think some HP calculators used this format.
4 bits were used to store a decimal digit (0 to 9). So, 8 bits could store 2 digits.
Of course, more bits are required to store a value but results are a lot easier to display.

[Tomorokoshi]

More from the originator of "Byte" with context of the times:

https://en.wikipedia.org/wiki/Werner_Buchholz

https://archive.org/stream/byte-magazine-1977-02/1977_02_BYTE_02-02_Usable_Systems#page/n145/mode/2up

[IDEngineer]

More from the originator of "Byte" with context of the times....

Very cool term he uses in that letter to the editor: "respelled". I'm adopting it into my personal lexicon!

This settles the (minor) question of whether the term "byte" could apply to anything other than a group of eight bits. And it's already been firmly established that "word" is similarly context-sensitive.

So is "kilo". As I said in my first post in this thread, context matters. Anyone with any experience in this industry won't misunderstand the numeric multiplier meant by the prefix kilo, based on context. I've never misunderstood it, I've never witnessed it being misunderstood, and I suspect the only people who "misunderstand" it do so for psuedo-nefarious purposes (read: Marketing) or pedanticism, like arguing about Oxford commas.

Let it go. "Kilo" is a non-problem.

[mikerj]

You will see a lot of "octets" when you start reading computer networking standards. Many of them date back to dinosaur era; some well-defined and unambiguous unit  was required to specify how data are to be transported over network and "byte" certainly wasn't that

When referring to ordering why is it always least/most significant byte rather than octet?  Is octet only used in a specific context?

[magic]

Depends on whom you ask.

If you look at the IPv4 spec, for example, they only talk about most significant bits and most significant octets.
Although figure 10 is titled "transmission order of bytes"

Most of the time no one cares about making a distinction because all general purpose computers have 8 bit bytes.
AFAIK in modern hardware problems only exist in some DSPs where the smallest unit of data might be 16 bits or more.

[SiliconWizard]

Long story short - KB (/MB/GB/...) is usually a non-issue in most cases given the appropriate context.

When it deals with binary *size*, it's a power of two multiplier in most cases (except in the storage industry...), and when it's about data throughput (such as in KB/s), it's already a lot more ambiguous and needs even more context or an appropriate definition (although again in the case of "K", there's this distinction with lowercase and uppercase, but there isn't for the higher multipliers... so that becomes ambiguous.)

I thus personally don't have a problem with the IEC standardization (even though I don't like the "kibi" term, but who cares, and btw, how do you spell MiB, GiB, ...? Is that MeBi, GiBi, ...?)

If there's even the slightest possibility of ambiguity, I think standardizing things is a sane idea. Just stating that "it's so obvious that it doesn't need that" is not very satisfying from an engineering POV.

[Nusa]

All this thread proves is that when a new "standard" is established, it takes more than 20 years for the general public to even hear about it, never mind follow it. Yes, that's how new those words are; they didn't exist before 1998 (I've been abusing computers since 1960). There are plenty of science units that have been around for centuries the general public never adopted. You just have to go with the flow and use context.