I often find myself explaining the same things in real life and online, so I recently started writing technical blog posts.
This one is about why it was a mistake to call 1024 bytes a kilobyte. It’s about a 20min read so thank you very much in advance if you find the time to read it.
Feedback is very much welcome. Thank you.
You must log in or register to comment.
A lot of people are replying as if OP asked a question. It’s a link to a blog post explaining why a kilobyte is 1000 and not 1024 bytes (exactly as the title says!). OP knows the answer, in fact they know it so well they wrote an extensive post about it.
Thank you for the write up! You should re-check the spelling and grammar as some sections had some troubles. I have a sentence I need to go to the post to get, so let me edit this later!
Edit: the second half of this sentence is a mess: “The factors don’t solely consist of twos, but ten are certainly lot of them.” Otherwise nothing jumped out at me but I would reread it just in case!
Thank you very much. I’ll try to fix that sentence later. I’m not a native speaker so it’s not always obvious for me when a sentence doesn’t sound right even though I pass sentences I’m not sure about through spell checks, MS Word grammar check and chat gpt 🤣
A lot of people are replying as if OP asked a question.
I think part of that is because outgoing links without a preview image are really easy to confuse with text-only posts, particularly because Reddit didn’t allow adding both a text and a link simultaneously. Though in this case the text should’ve tipped people off that there’s a link as well.
As for the actual topic, I agree with OP. I often forget to do it right when speaking, but I try to at least get it right when writing.
OP asked for feedback.
Well it’s because computer science has been around for 60+ years and computers are binary machines. It was natural for everything to be base 2. The most infuriating part is why drive manufacturers arbitrarily started calling 1000 bytes a kilobyte, 1000 kilobytes a megabyte, and 1000 megabytes a gigabyte, and a 1000 gigabytes a terabyte when until then a 1 TB was 1099511627776 bytes. They did this simply because it made their drives appear 10% bigger. So good ol’ shrinkflation. You could make drives 10% smaller and sell them for the same price.
If a hard drive has exactly 8’269’642’989’568 bytes what’s the benefit of using binary prefixes instead of decimal prefixes?
There is a reason for memory like caches, buffer sizes and RAM. But we don’t count printer paper with binary prefixes because the printer communication uses binary.
There is no(!) reason to label hard drive sizes with binary prefixes.
Pretty obvious that you didn’t read the article. If you find the time I’d like to encourage you to read it. I hope it clears up some misconceptions and make things clearer why even in those 60+ years it was always intellectually dishonest to call 1024 byte a kilobyte.
You should at least read “(Un)lucky coincidence”
Ok so I did read the article. For one I can’t take an article seriously that is using memes. Thing the second yes drive manufacturers are at fault because I’ve been in IT a very very long time and I remember when HD manufacturers actually changed. And the reason was greed (shrinkflation). I mean why change, why inject confusion where there wasn’t any before. Find the simplest least complex reason and that is likely true (Occam’s razor). Or follow the money usually works too.
It was never intellectually dishonest to call it a kilobyte, it was convenient and was close enough. It’s what I would have done and it was obviously accepted by lots of really smart people back then so it stuck. If there was ever any confusion it’s by people who created the confusion by creating the alternative (see above).
If you wanna be upset you should be upset at the gibi, kibi, tebi nonsense that we have to deal with now because of said confusion (see above). I can tell you for a fact that no one in my professional IT career of over 30 years has ever used any of the **bi words.
You can be upset if you want but it is never really a problem for folks like me.
Hopefully this helps…
I just think that kilobyte should have been 1000 (in binary, so 16 in decimal) bytes and so on. Just keep everything relating to the binary storage in binary. That couldn’t ever become confusing, right?
Because your byte is 10 decimal bits, right? EDIT: Bit is actually an abbreviation, BIT, initially, so it would be what, DIT?.. Dits?..
kilobit = 1000 bits. Kilobyte = 1000 bytes.
How is anything about that intellectually dishonest??
The only ones being dishonest are the drive manufacturers, like the person above said. They sell storage drives by advertising them in the byte quantity but they’re actually in the bit quantity.
They sell storage drives by advertising them in the byte quantity but they’re actually in the bit quantity.
No, they absolutely don’t. That’d be off by 8x.
The subject at hand has nothing to do with bits. Please, read what OP posted. It’s about 1024 vs 1000
Calling 1024 a kilo is intellectually dishonest. Your conversation is perfectly fine.
Thanks for this article. Unfortunately, you used the word “prefix” when you really meant “unit symbol”. So, “kilo” and “mega” are prefixes, kB and MB are unit symbols. You repeatedly called the latter “prefixes”.
Thank you for the feedback. I know that only the “first” part is the prefix and I tried to be careful to not use it wrong. I just checked all 53 instances of “prefix” and I don’t see a wrong one, but to be fair there are situations that could be misunderstood easily like here:
Today the only correct conversions are to either use SI prefixes (like 1 MB = 1000² bytes) or binary prefixes (1 MiB = 1024² bytes).
But with prefix I only meant the “M” and “Mi” part and they are both prefixes.
I’ll try to clarify that later so the difference is clear to all readers. Thank you.
Ok, I understand what you are trying to do, but I that is not how I read it at the time. Prefix to me in this context means e.g., “kilo” in “kilobyte”, and not the “k” in “kB”. I am not sure it is helpful to split the unit symbol up like that.
In terms of language you are correct. But in terms of SI usage it seems to me OP is expressing it correctly. The SI unit prefixes have a name, a symbol and a multiplier. The prefix is a concept that encompasses all three of those attributes. So “kilo” is one way of identifying the 10^3 unit prefix, but the name kilo is not the prefix itself. It’s just the name we use to refer to it. And the symbol k in km is certainly the unit prefix portion of that unit of measure.
But the first part is called prefix even in the standard itself. I wanted to make that distinction because it’s not important what the base unit is. By speaking about prefixes instead of the unit as a whole I wanted to make it clear that you can (at least in theory) use any base unit. So everything I said about KiB and kB is also true for Kib and kb and even for kK (kilokelvin) and KiB (kibikelvin) 🤣
While we’re nitpicking, the post says multiple times that SI prefix symbols are “all uppercase except for kilo (k)”.
That’s just factually wrong. More than half of them are lowercase! There’s centi- ©, micro- (µ), nano- (n), etc. On the positive side there’s even deca- (da) and hecto- (h), though they aren’t particularly common or useful. I did at least see milli- (m) and bit (b) mentioned in a brief note though.
Obviously context matters and only the positive powers from kilo upward are relevant in computer science. But I studied chemistry and physics so I guess it irked me to see the statement repeatedly ignore all the negative powers of ten.
Overall, good rant though 😅 I’ll be more careful to use KiB and MiB from here out when appropriate.
❤️ Thank you for taking the time to read it. And thank you so much for pointing that out, you are completely right and I totally didn’t think about that while writing the article, probably because negative exponents are pretty rare in computer science (as in milli-bytes, etc.). I’ll fix that in a few days. Thanks again for pointing that out.
Here’s my favorite part.
“In addition, the conversions were sometimes not even self-consistent and applied completely arbitrary. The 3½-inch floppy disk for example, which was marketed as “1.44 MB”, was actually not 1.44 MB and also not 1.44 MiB. The size of the double-sided, high-density 3½-inch floppy was 512 bytes per sector, 18 sectors per track, 160 tracks, that’s 512×18×16 = 1’474’560 bytes. To get to “1.44” you must first divide 1’474’560 by 1024 (“bEcAuSE BiNaRY obviously”) to get 1440 and then divide by 1000 for perfect inconsistency, because dividing by 1024 again would get you an ugly number and we definitely don’t want that. We finally end up with “1.44”. Now let’s add “MB” because why the heck not. We already abused those units so much it’s not like they still mean anything and it’s “close enough” anyways. By the way, that “close enough” excuse never “worked when I was in school but what would I know compared to the computer “scientists” back then.
When things get that messy, numbers don’t even mean anything any more. Might as well just label the products using entirely qualitative terms like “big” or “bigger”.
❤️ Thank you for taking the time to read it.
I was confused when I just read the headline. Should be “Why I (that would be you not me) think a kilobyte should be 1000 instead of 1024”. Unpopular opinion would be a better sub for it.
You should read the blog post. It’s not a matter of option.
Just because you wrote about a topic doesn’t mean you’re suddenly the authority figure lol.
I know there is no option as 1024 is what the standard is now. Im not reading that anymore than someone saying how a red light really means go.
1024 is not the standard. The standard term for 1024 is “kibi” or “Ki” and the standard term for 1000 is “kilo” and has been since the year 1795.
There was a convention to use kilo for 1024 in the early days of computing since the “kibi” term didn’t exist until 1998 (and took a while to become commonly used) — but that convention was always recognised as an incorrect use of the term. People just didn’t care much especially since kilobytes were commonly rounded anyway. A 30,424 byte file is 29.7109375 kibibytes or 30.424 kilobytes… both will likely be rounded to 30 either way, so who cares if it’s slightly wrong? Just use bytes if you need to know the exact size.
Also - hard drives, floppy disks, etc have always referred to their size in base 1000 numbers so if you were working with 30KB in the early days of computers it was very rarely RAM. A PDP-11 computer, for example, might have only had 8196 bytes of RAM (that’s 8 kibibytes).
There are some places where the convention is still used and it can be pretty misleading as you work with larger numbers. For example 128 gigs equals 128,000,000,000 bytes (if using the correct 1000 unit) or 137,438,953,472 bytes (if kilo/mega/giga = 1024).
The “wrong” convention is commonly still used for RAM chips. So a 128GB RAM chip is significantly larger than a 128GB SSD.
Also - hard drives, floppy disks, etc have always referred to their size in base 1000 numbers
That is not true. For a long time everything (computer related) was in the base 2 variants. Then the HD manufacturers changed so their drives would appear larger than they actually were (according to everyone’s notions of what kn/mb/gb meant). It was a marketing shrinkflation stunt.
Removed by mod
I’m not sure if I’m too stupid, but how so?
Removed by mod
So why don’t they just label drives in Terabit instead of terabyte. The number would be even bigger. Why don’t Europeans also use Fahrenheit, with the bigger numbers the temperature for sure would instantly feel warmer 🤣
Jokes aside. Even if HDD manufacturers benefit from “the bigger numbers” using the 1000 conversation is the objectively only correct answer here, because there is nothing intrinsically base 2 about hard drives. You should give the blog post a read 😉
there is nothing intrinsically base 2 about hard drives
Yes there is. The addressing protocol. Sectors are 512 (2⁹) bytes, and there’s an integer number of them on a drive.
That’s true but the entire disk size is not an exact power of two that’s why binary prefixes (1024 conversation) don’t have any benefit whatsoever when it comes to hard drives. With memory it’s a bit different because other than with storage devices RAM size is always exactly a power of two.
The mistake is thinking that a 1000 byte file takes up a 1000 bytes on any storage medium. The mistake is thinking that it even matters if a kB means 1000 or 1024 bytes. It only matters for some programmers, and to those 1024 is the number that matters.
Disregarding reality in favor of pedantics is the real mistake.
I dunno it makes up a few gigabytes of lost storage on a terrabyte hard drive.
This whole mess regularly frustrates me… why the units can’t be used consistently?!
The other peeve of mine with this debacle is that drive capacities using SI units do not use the full available address space (since it’s binary). Is the difference between 250GB and 256GiB really used effectively for wear-levelling (which only applies to SSDs) or spare sectors?
Power of 2 makes more sense to the computer. 1000 makes more sense to people.
Of course. The thing is, though, that if the units had been consistent to begin with, there wouldn’t be anywhere near as much confusion. Most people would just accept MiB, GiB, etc. as the units on their storage devices. People already accept weird values for DVDs (~4.37GiB / 4.7GB), so if we had to use SI units then a 256GiB drive could be marketed as a ~275GB drive (obviously with the non-rounded value in the fine print, e.g. “Usable space approx. 274.8GB”).
They were consistent until around 2005 (it’s an estimate) when drives got large enough where the absolute difference between the two forms became significant. Before that everyone is computing used base 2 prefixes.
I bet OP does too when talking about RAM.
It’s not as simple as that. A lot of “computer things” are not exact powers of two. A prominent example would be HDDs.
In terms of storage 1000 and 1024 take the same amount of
bytesbits to represent. So from a computer point of view 1024 makes a lot more sense.It’s just a binary Vs decimal thing. 1000 is not nicely represented in binary the same as 1024 isn’t in decimal.
Edit: was talking about storing the actual number.
In terms of storage 1000 and 1024 take the same amount of bytes.
What? No. A terabyte in 1024 units is 8,796,093,022,208 bits. In 1000 units it’s 8,000,000,000,000 bits.
The difference is substantial with larger numbers.
Both require the same amount of bits again. So the second one makes more sense for a computer.
Huh? What does how a drive size is measured affect the available address space used at all? Drives are broken up into blocks, and each block is addressable. This is irrelevant of if you measure it in GB or GiB and does not change the address or block size. Hell, you have have a block size in binary units and the overall capacity in SI units and it does not matter - that is how it is typically done with typical block sizes being 512 bytes, or 4096 (4KiB).
Or have anything to do with ware leveling at all? If you buy a 250GB SSD then you will be able to write 250GB to it - it will have some hidden capacity for ware-leveling, but that could be 10GB, 20GB, 50GB or any number they want. No relation to unit conversions at all.
Huh? What does how a drive size is measured affect the available address space used at all? Drives are broken up into blocks, and each block is addressable.
Sorry, I probably wasn’t clear. You’re right that the units don’t affect how the address space is used. My peeve is that because of marketing targeting nice round numbers, you end up with (for example) a 250GB drive that does not use the full address space available (since you necessarily have to address to up 256GiB). If the units had been consistent from the get-go, then I suspect the average drive would have just a bit more usable space available by default.
My comment re wear-levelling was more to suggest that I didn’t think the unused address space (in my example of 250GB vs 256GiB) could be excused by saying it was taken up by spare sectors.
(for example) a 250GB drive that does not use the full address space available
Current drives do not have different sized addressable spaces and a 256GiB drive does not use the full address space available. If it did then that would be the maximum size a drive could be. Yet we have 20TB+ drives and even those are no where near the address size limit of storage media.
then I suspect the average drive would have just a bit more usable space available by default.
The platter size might differ to get the same density and the costs would also likely be different. Likely resulting in a similar cost per GB, which is the number that generally matters more.
My comment re wear-levelling was more to suggest that I didn’t think the unused address space (in my example of 250GB vs 256GiB) could be excused by saying it was taken up by spare sectors.
There is a lot of unused address space - there is no need to come up with an excuse for it. It does not matter what size the drive is they all use the same number of bits for addressing the data.
Address space is basically free, so not using it all does not matter. Putting in extra storage that can use the space does cost however. So there is no real relation between the address spaces and what space is on a drive and what space is accessible to the end user. So it makes no difference in what units you use to market the drives on.
Instead the marketing has been incredibly consistent - way back to the early days. Physical storage has essentially always been labeled in SI units. There really is no marketing conspiracy here. It just that is they way it was always done. And why it was picked that way to begin with? Well, that was back in the day when binary units where not as common and physical storage never really fit the doubling pattern like other components like ram. You see all sorts of random sizes in early storage media so SI units I guess did not feel out of place.
You know what else is frustrating? Time units. It’s like we’re back in the pre-SI days again. Try to compare the flow rates of two pumps when one is 123 m^3/h and the other is 1800 l/min. The French tried to fix this mess too while they were at it, but somehow we’re still stuck with this archaic mess.
The other peeve of mine with this debacle is that drive capacities using SI units do not use the full available address space (since it’s binary).
The “full available address space” goes down as the drive gets older and bad sectors are removed.
With a good drive, it might take ten or more years before you actually see the “size” of the drive shrink, but that’s mostly because you 500GB drive actually had something like 650GB of storage when it was brand new.
This has been my pet rant for a long time, but I usually explain it … almost exactly the other way around to you.
You can essentially start off with nothing using binary prefixes. IBM’s first magnetic harddrive (the IBM 350 - you’ve probably seen it in the famous “forklifting it into a plane” photo) stored 5 million characters. Not 5*1024*1024 characters, 5,000,000 characters. This isn’t some consumer-era marketing trick - this is 1956, when companies were paying half a million dollars a year (2023-inflated-adjusted) to lease a computer. I keep getting told this is some modern trick - doesn’t it blow your mind to realise hdd manufacturers have been using base10 for nearly 70 years? Line-speed was always
a liebase 10, where 1200 baud laughs at your 2^n fetish (and for that matter, baud comes from telegraphs, and was defined before computers existed), 100Mbit ethernet runs on a 25MHz clock, and speaking of clocks - kHz, MHz, MT/s, GT/s etc are always specified in base 10. For some reason no-one asks how we got 3GHz in between 2 & 4GHz CPUs.As you say, memory is the trouble-maker. RAM has two interesting properties for this discussion. One is that it heavily favours binary-prefixed “round numbers”, traditionally because no-one wanted RAM with un-used addresses because it made address decoding nightmarish (tl;dr; when 8k of RAM was usually 8x1k chips, you’d use the first 3 bits of the address to select the chip, and the other 10 bits as the address on the chip - if chips didn’t use their entire address space you’d need to actually calculate the address map, and this calculation would have to run multiples of times faster than the cpu itself) . The second, is that RAM was the first place non-CSy types saw numbers big enough for k to start becoming useful. So for the entire generation that started on microcomputers rather than big iron, memory-flavoured-k were the first k they ever tasted.
I mean, hands up who had a computer with 8-64k of RAM and a cassette deck. You didn’t measure the size of your stored program in kB, but in seconds of tape.
This shortcut than leaked into filesystems purely as an implementation detail - reading disk blocks into memory is much easier if you’re putting square pegs into square holes. So disk sectors are specified in binary sizes to enable them to fit efficiently into memory regions/pages. For example, CP/M has a 128-byte disk buffer between 0x080 and 0x100 - and its filesystem uses 128-byte sectors. Not a coincidence.
This is where we start getting into stuff like floppy disk sizes being utter madness. 360k & 720k were 720 and 1440 512-byte sectors. When they doubled up again, we doubled 2800 512-byte sectors gave us 1440k - and because nothing is ever allowed to make sense (or because 1.40625M looks stupid), we used base10 to call this 1.44M.
So it’s never been that computers used 1024-shaped-k’s. It should be a simple story of “everything uses 1,000s, except memory because reasons”. But once we started dividing base10-flavoured storage devices into base2-flavoured sectors, we lost any hope of this ever looking logical.
aside: the little-k thing. SI has a beautifully simple rule, capital letters for prefixes >1, small letters for prefixes <1. So this disambiguates between a millivolts (mV) and megavolts (MV).
But, and there’s always a but. The kilogram was the first SI unit, before they’d really thought it through. So we got both a lower-case k breaking such a beautifully simple rule, and the kilogram as a base unit instead of a gram. The Kilogram is metric’s “screw it, we’ll do it live”.
Luckily this is almost a non-issue in computing as a fraction of a bit never shows up in practice. But! If you had a system that took 1000 seconds to transfer one bit, you could call that a millibit per second, or mbps, and really mess things up.
2, 4, 8, 16, 32, 64, 128, 256, 512, 1024. It’s pretty fucking logical m8. You know what’s not logical? Base 10
Yeah, base ten really screws around with programming. You specifically have to use a decimal type if you really want to use it (for like finance or something), but it’s much slower.
I know it’s already been explained but here is a visualization of why.
0 2 4 8 16 32 64 128 256 512 1024
Did you read the blog post? If you don’t find the time you should at least read “(Un)lucky coincidence” to see why it’s not (and never was) a bright idea to call 1024 “a kilo”.
Dude you’re pretty condescending for a new author on an old topic.
Yeah I read it and it’s very over worded.
1024 was the closest binary approximation of 1000 so that became the standard measurement. Then drive manufacturers decided to start using decimal for capacity because it was a great way to make numbers look better.
Then the IEC decided “enough of this confusion” and created binary naming standards (kibi gibi etc…) and enforced the standard decimal quantity values for standard names like kilo-.
It’s not ground breaking news and your constant arguing with people in the thread paints you as quite immature. Especially when plenty of us remember the whole story BECAUSE WE LIVED IT AS IT PROFESSIONALS.
We lacked a standard, a system was created. It was later changed to match global standard values.
You portray it with emotive language making decisions out to be stupid, or malicious. A decision was made that was perfectly sensible at the time. It was then improved. Some people have trouble with change.
Your writing and engagement styles scream of someone raised on clickbait news. Focus on facts, not emotion and sensationalism if you want to be taken seriously in tech writing.
Focus on emotion and bullshit of you want to work for BuzzFeed.
And if you just want an argument go use bloody twitter.
Because SI prefixes are always powers of the base. Base 10 is the most common, but that’s more human psychology that math.
SI prefixes are literally just base ten and not really about human psychology.
I think they mean it’s easier to refer to powers of 1000 with the SI units, rather than of 1024 as with Kibi and the lot. Especially higher up in the prefixes, because it starts to diverge more and more from the expected value.
Please count your fingers and reevaluate your response.
It’s a scam by HDD makers to sell less storage for more money.
that’s what it was initially, reporting decimal ‘megabytes’ for hdd capacity. lawsuits and settlements followed.
the dust settled and what we have now is disclaimers on storage products (from the legal settlements) and they continue to use ‘decimal’ measurements…
and we also a different set of prefixes for ‘binary’ units of measurements (standards body trying to address the problem of confusion): kibi, mebi, gibi, tebi, pebi, exbi; which are not widely used yet… the ‘old’ ones are for decimal but still commonly used for binary.
Did you read the blog post? It’s not a scam. HDD vendors might profit from “bigger numbers” but using the units they do is objectively the only sensible and correct option. It’s like saying that the weather report is in Fahrenheit because in Celsius the numbers would be lower and feel somehow colder 🤣
If it would be about bigger numbers why don’t HDD manufacturers just use Terabit instead of terabyte? The “bigger number” argument is not a good one.
Because it’s much easier to mistake a number for a somewhat close number than one that is orders of magnitude different…
I’ll try to read the article later but the reality is that HDD manufacturers could help customers disambiguate but that would hurt their bottom line so they don’t.
Videogame companies literally did use “megabit” when the truth was “128KiB”, because it sounded better. Actual computer companies were still listing binary power numbers, because buyers had more to invest and care about accuracy.
You say “sensible”, but it’s lying for profit.
It is only a mistake from a Human PoV. It is more efficient for the chip since 1000 bytes and 1024 bytes take up the same space. But Humans find anything not base 10 difficult.
It’s not really about the space numbers need inside the computer but about unit prefixes.
Which are the references to drive space. What is your point?
Nice to learn about the SI standard notation KiB, MiB, etc. I had no idea.
KiB and MiB are not SI prefixes but IEC binary prefixes but the names are derived from the SI names for simplicity.
I suggest considering this from a linguistic perspective rather than a technical perspective.
For years (decades, even), KB, MB, GB, etc. were broadly used to mean 2^10, 2^20, 2^30, etc. Throughout the 80s and 90s, the only place you would likely see base-10 units was in marketing materials, such as those for storage media and modems. Mac OS exclusively used base-2 definitions well into the 21st century. Windows, as noted in the article, still does. Many Unix/POSIX tools do, as well, and this is unlikely to change.
I will spare you my full rant on the evils of linguistic prescriptivism. Suffice it to say that I am a born-again descriptivist, fully recovered from my past affliction.
From a descriptivist perspective, the only accurate way to define kilobyte, megabyte, etc. is to say that there are two common usages. This is what you will see if you look up the words in any decent dictionary. e.g.:
- https://www.dictionary.com/browse/kilobyte
- https://www.merriam-webster.com/dictionary/kilobyte
- https://en.wiktionary.org/wiki/kilobyte
I don’t recall ever seeing KiB/MiB/etc. in the 90s, although Wikipedia tells me they “were defined in 1999 by the International Electrotechnical Commission (IEC), in the IEC 60027-2 standard”.
While I wholeheartedly agree with the goal of eliminating ambiguity, I am frustrated with the half-measure of introducing unambiguous terms on one side (KiB, MiB, etc.) while failing to do the same on the other. The introduction of new terms has no bearing on the common usage of old terms. The correct thing to have done would have been to introduce two new unambiguous terms, with the goal of retiring KB/MB/etc. from common usage entirely. If we had KiB and KeB, there’d be no ambiguity. KB will always have ambiguity because that’s language, baby! regardless of any prescriptivist’s opinion on the matter.
Sadly, even that would do nothing to solve the use of common single-letter abbreviations. For example, Linux’s
ls -l -hcommand will return sizes like 1K, 1M, 1G, referring to the base-2 definitions. Only if you specify the non-default--siflag will you receive base-10 values (again with just the first letter!). Many other standard tools have no such options and will exclusively use base-2 numbers.
