Wikipedia talk:Manual of Style (dates and numbers)/Archive 65

From Wikipedia, the free encyclopedia
Jump to: navigation, search

The below discussion titled "Converting to SI prefixes" also continues onto Archive 66.

Converting to SI prefixes

I'm presently dealing with a new user whose sole contribution to the encyclopedia has been to push SI prefixes into a variety of Macintosh hardware articles. And by "dealing with", I mean, reverting all their changes.

There's a serious problem here with the stated MOS guideline: It contravenes both WP:NPOV and WP:V. In essence, SI prefixes are meant to solve an ambiguity issue between "powers of 2" and "multiples of 10". This decision was made by a task force a few years ago, and was adopted by various organisations and communities as a result. The problem is that most major manufacturers of hardware and software, as well as most publications and news sources, have not adopted this naming for whatever reason. I don't necessarily agree with it, but that's their collective perogative.

So what's an encycopedia to do? The answer seems clear enough: our core policies revolve around a neutral presentation of our sources, which means it behooves us to use MB, GB, etc. when our sources use those prefixes. Wikipedians should absolutely not take it upon themselves to state numbers differently from how our verifiable, reliables sources do. Arguments along the lines of "But Apple does it wrong!" don't wash either, since that's a matter of personal opinion. Using SI prefixes to describe consumer hardware is nothing more than a minority viewpoint at this point. Personally, I'd like to see these prefixes be adopted more widely so as to solve this ambiguity, but Wikipedia isn't the place to champion that agenda. We report on what others say, and that's it... policies always override MOS guidelines. -/- Warren 22:31, 21 January 2007 (UTC)

I'm afraid I disagree-we need not always report sources verbatim. Apple may call their new laptop "the best thing to happen to graphics designers since Photoshop", but that doesn't mean we've got to put that verbatim in any article which mentions it! Similarly, KiB/MiB/GiB are the correct ways to represent binary values. The prefix "kilo" means "one thousand". That is not a matter of opinion, any more than it's a "matter of opinion" to say that three is a quantity one greater than two and one less than four. That is not my opinion, that's what the word means. Similarly, if a source is using imprecise and inaccurate terminology, especially when their terminology is intended to appeal to a mass audience, we need not repeat it exactly-in our context, so long as the meaning is clear, we can certainly use better, more precise terminology. In this case, it is indisputably clear that the numbers referred to are binary values. In this case, the use of XiB prefixes are correct, and the use of XB prefixes are incorrect-we're not referring to the numbers that "kilo", "mega", and the like refer to! That is not an opinion, it is verifiably and provably true. Seraphimblade 00:54, 22 January 2007 (UTC)
You're advocating silently correcting our sources for them when they're wrong, then? Ehhhh, something seems really wrong with that. I'd rather let articles on megabyte etc. explain the issue (per WP:NPOV and Jimbo's own assessment that minority viewpoints should be in ancilliary articles -- YES, the "correct" measurement is a matter of opinion)... or specifically disclaim the conversion on pages where we use it. We don't generally use metric measurements on articles on American subjects, do we? Nobody would advocate switching from miles to kilometers in those Wikipedia just because NIST said so, would they? The answer is no -- we tend to choose what to present based on common usage in the article's domain space, not based on what governments and standards organisations (or advocates thereof) say we should use. -/- Warren 01:21, 22 January 2007 (UTC)
However, here, we're dealing with a layman's term which is commonly used, but is also factually and provably incorrect. Whether to say "one inch" or "2.54 centimeters" is indeed a matter of opinion-either one is correct. However, if one of our sources was incorrectly using a conversion factor of 1 inch = 3.54 centimeters, they are wrong. It's not my opinion, in that case, that they're wrong. They're wrong. I would agree when directly quoting, we certainly should not change a direct quote (though we probably should tag it with (sic) where the incorrect terminology appears). But when paraphrasing, yes, we should put an obviously-incorrect layman's term into the correct technical term. The prefix kilo means 1000. That is a fact. "Kilobyte" means 1000 bytes, just like "kilometer" means 1000 meters and "kilogram" means 1000 gram. That's just what the prefix means. If a source uses "kilometer" to refer to a distance which is actually 1024 meters, they're wrong. If a source uses "kilobyte" to refer to 1024 bytes, they're wrong. That's not a matter of opinion. They may have figured it for "close enough", in what they were doing, but we should be as precise as possible. Especially up into the MiB/GiB ranges, there is a significant variance between the base 10 and binary prefixes, and what kind of number they indicate. This is not a matter of opinion, however. Either there's 1000 bytes, or there's 1024. In the cases we're talking about here, there are 1024. KiB properly represents that. KB does not. That's not my opinion, or anyone's. Words have meanings, and so do prefixes. K means 1000. Ki means 1024.
I suppose our other option would be to simply eschew prefixes and report all values in bytes, but that would get cumbersome, to say the least. However, just because most computer manufacturers would rather settle for "close enough" and allow error doesn't mean we need to, or should. That is not a "rewording" of what the source said, any more than any paraphrase is. But this is no "minority viewpoint"-find me any computer scientist who will disagree on the actual binary values. For that matter, not even Apple or Dell would disagree-they're just using incorrect but widely-recognized terms for marketing purposes. The "minority viewpoint" criteria do not apply when the viewpoint in question is effectively universal! Seraphimblade 01:35, 22 January 2007 (UTC)
No, kibibit, etc. are ill-conceived needless committee-driven nonsense. The English language is quite able to to accommodate words with slightly different meanings dependent on context. —Centrxtalk • 01:45, 22 January 2007 (UTC)
Not quite necessarily-the reason I advocate use of precision numbers was, first, when I started getting angry calls from people whose systems I'd upgraded. I promised them a 200 GB drive, you see, but it says it's only 185! It generally took a good deal of explaining to get them to comprehend they hadn't gotten shortchanged, and even then some of them seemed rather unconvinced. (Anymore, I explain this beforehand, and people don't seem to have much trouble understanding the basics-the "gigabyte" term is ambiguous, and the file system requires some space to set up its tables and such.) At least I never got sued over the discrepancy, though-but it sure looks like a lot of people have! There is not a thing wrong with being precise or taking proactive steps to prevent confusion, and relying on "context" to handle such things inevitably leads to misinformation and misunderstanding sooner or later. Perhaps, in most cases, context would handle it fine. But when a better, more correct, more precise alternative is available, why not use it? Seraphimblade 02:10, 22 January 2007 (UTC)
Actually, it is incorrect to say that one kilobyte can only mean 1000 bytes, at least in American English. See, for example, Merriam-Webster's definition of "kilobyte" and Free On-Line Dictionary of Computing's definition of "kilobyte". Even dictionaries which include the smaller variant list that as a second, less common definition; see, for example,'s definitions of "kilobyte". --ΨΦorg 04:55, 22 January 2007 (UTC)
Interesting. Given that, staying with xB might make sense, provided we take care to link to something that clearly explains the discrepancy. (Still, I fail to see what harm it would do to use more precise terminology when it's known which one applies.) Seraphimblade 06:59, 22 January 2007 (UTC)
I totally agree with Seraphimblade. This is not a matter of opinion. Even if dictionaries say that kilobyte could mean 1024 bytes, it is ambiguous and an encyclopedia should be precise and consistent. There are prefixes specially designed for this purpose. They are internationally adopted as standards and they are unambiguous : what is written is what is meant, why not use it ? Sarenne 11:38, 22 January 2007 (UTC)

We can decide to use whatever convention we like, but we should be consistent. If it's clear that a source is using MB to mean 2^20 bytes, there is no problem with us reporting it as MiB, if that's the convention we decide on. The only problem is when it's not clear what the source means, in which case I guess the best we can do is use whatever they use. --Tango 11:50, 22 January 2007 (UTC)

I would agree, if there could be any doubt as to which one the source actually means, we should stay with what they use. In most cases, though, it's pretty clear whether it's a reference to base-10 or base-2. Seraphimblade 11:53, 22 January 2007 (UTC)
I agree, but the question is: Do we stay consistent with our sources and with common usage in the article domain? Or do we stay cosnistent across the encyclopedia? When this question has been asked of other measurement and naming standards, we have almost always decided on "sources and common usage". We use British spelling and linguistics in British articles; we use kilometers in Canadian articles; we use the time zone relevant to an event's location; we have also had a strong tendency towards binary prefixes in popular computing articles, because that's what's used in general conversation, print and online media, and by the manufacturers themselves. That's a pretty large group of real people to turn our backs on and say, "no, you're wrong, these technical committees know better". The article on megabyte does explain the ambiguity early on.
Hmm... a footnote or some other kind of explanatory widget in articles where this is an issue may be a good answer here. We do this to resolve ambiguity in other areas (pronunciation, for example.) Heck, I'm willing to do that work myself in the 1,500+ computing articles I have on my watchlist where these measurements would be relevant. -/- Warren 13:53, 22 January 2007 (UTC)
A footnote or something like that might work. Maybe some kind of template could be devised? I'm not going for any single solution, I just think we should make the issue clear to our readers-a lot of people have no idea there even is any kind of discrepancy. Seraphimblade 14:00, 22 January 2007 (UTC)
I think we should go by what would be the world wide norm, even apple uses MB and GB not MiB or GiB,No where in my computer does it reference MiB or GiB I have visted every major computer site and they all use the MB and GB term.Not one of them uses the GiB or MiB term that I saw.  Planetary Chaos  Talk to me  17:52, 22 January 2007 (UTC)

How ever I do see that if you have a 512 memory it would correctly be 512 MiB but if you have a 1 GB it would not be a GiB.  Planetary Chaos  Talk to me  17:56, 22 January 2007 (UTC)

You are still confusing prefixes... If you have a "1 GB" memory, you have 2 times a "512 MB" memory. A "512 MB" memory is actually a 512 MiB memory, so you have 2*512 MiB = 1024 MiB = 1 GiB. Sarenne 18:07, 22 January 2007 (UTC)
  • one mebibyte (MiB) is 1048576 bytes

one Megabyte (MB) is 1000000 bytes Source I don't think they are wrong.[1]  Planetary Chaos  Talk to me  18:30, 22 January 2007 (UTC)

Hard drive seller "gigabytes" are neither binary gigabytes nor decimal gigabytes. They are usually either 1,000,000 binary kilobytes (i.e. 1,024,000,000 bytes) or 1,000 binary megabytes (i.e. 1,048,576,000 bytes). It's weird how they confuse and conflate the two meanings into a single prefix. --ΨΦorg 18:32, 22 January 2007 (UTC)
No, Hard Drive sellers are using "decimal" gigabytes (1,000,000,000 bytes) and "decimal" megabytes Example. They are using the prefixes like they should be used. Sarenne 18:47, 22 January 2007 (UTC)
Why do you quote this source ? I'm telling you that a 1 GB RAM has an actual capacity of 1 GiB, nothing more.Sarenne 18:47, 22 January 2007 (UTC)

All i'm saying is that MiB and GiG are not part of the measuring system in place now.and untill this changes, what's in place now should be reflected in the Wikipedia articles.

"It is important to recognize that the new prefixes for binary multiples are not part of the International System of Units (SI), the modern metric system." "Faced with this reality, the Institute of Electrical and Electronics Engineers ( IEEE ) Standards Board decided that IEEE standards will use the conventional, internationally adopted, definitions of the SI prefixes. Mega will mean 1 000 000, except that the base-two definition may be used (if such usage is explicitly pointed out on a case-by-case basis) until such time that prefixes for binary multiples are adopted by an appropriate standards body. "  Planetary Chaos  Talk to me  19:19, 22 January 2007 (UTC)

If you want to use SI prefixes with their real values (why not...), you have to change the numbers. When Apple, MS, Intel or AMD says 512 MB, it means 512 MiB so if you want to keep the real MB unit (=1 000 000 bytes) you have to change 512 MB to 536.870912 MB. Isn't it easier to say 512 MiB than 536.870912 MB ? Sarenne 19:45, 22 January 2007 (UTC)

No, I think it's easier to stick with the accepted (world wide)values instead of the confusing "new" prefixes that are not yet recognized by the world community.  Planetary Chaos  Talk to me  20:02, 22 January 2007 (UTC)

So why do you quote the IEEE that says the opposite ? What is the accepted value of 1 MB according to you ? Sarenne 20:13, 22 January 2007 (UTC)

as quoted above Mega will mean 1 000 000.You might want to read the IEEE currant standards.  Planetary Chaos  Talk to me  21:01, 22 January 2007 (UTC)

The "generally accepted values" are very dependant on context, and even then they aren't always universal. For example, should the size of my hard drive be reported as "200GB" as the manufacturer reports it, or "189GB" as Windows reports it? (Interestingly, 189GiB=203GB, I guess the 200GB it was sold as was approximate...) --Tango 21:03, 22 January 2007 (UTC)

IMO (and that's also what the current MoS recommends), the size of your hard drive should be reported as "200 GB" because 1) It uses the standard SI prefix G=109 2) It is consistent with the use of binary prefixes for binary capacities 3) Manufacturers chose this unit so there is no conversion to do.
Planetary Chaos, if 1 000 000 bytes is the value of 1 MB and you don't want to use binary prefixes, you are aware that it implies that we change all the capacity values when dealing with memory capacities (512 MB becomes 536.870912 MB, 1 GB becomes 1,073741824 GB...)? Is it really what you want to do ? Sarenne 21:27, 22 January 2007 (UTC)
I don't understand your point (2), using 200GB isn't using binary prefixes at all... And point (3) doesn't say why the manufacturer's choice should trump Microsoft's choice. Anyway, it was a rhetorical question - I was just pointing out that is isn't as simple as saying "Use the generally accepted values". --Tango 23:26, 22 January 2007 (UTC)
Using "200 GB" isn't using binary prefixes but is coherent with the use of binary prefixes. You can use "200 GB" for hard drive and GiB for RAM in the same article. If you use "189 GB" for the HD and "1 GiB" for the RAM, it is not consistent since in that case GiB and GB mean the same thing. Sarenne 23:49, 22 January 2007 (UTC)

Tango,Is your hard drive shared? Meaning your hard drive is typically C do you have a restore D or E partitioned?  Planetary Chaos  Talk to me  21:44, 22 January 2007 (UTC)

The hard drive in question is an external drive (well, a normal drive in an external caddy, to be precise), and is unpartitioned (or has a single partition, to be precise). --Tango 23:26, 22 January 2007 (UTC)

Sarenne,it is not what I want,but it is however the international standard.  Planetary Chaos  Talk to me  22:01, 22 January 2007 (UTC)

Binary prefixes (Ki, Mi, Gi...) are international standards too. Sarenne 22:17, 22 January 2007 (UTC)

Show your source.  Planetary Chaos  Talk to me  22:24, 22 January 2007 (UTC)

Binary prefix, IEEE 1541, IEC 60027, BIPM CCU, BIPM prefixes, ANSI. Sarenne 22:38, 22 January 2007 (UTC)

Let me say this M as in Mega has allways been a million long before the introduction of the PC.I to have a 200 GB HD 195 GB on C and 5 GB on RECOVERY D .Now correct me if my math is not correct but 195 + 5 = 200?  Planetary Chaos  Talk to me  00:04, 23 January 2007 (UTC)

Goody, this biscuit again... Since I haven't seen any arguments that haven't been used before, I'll just say that I don't see any compelling reason to change the current style guidelines on this matter. Conforming strictly to the SI definitions of the prefixes in question and using the binary prefixes proposed by IEC (and endorsed by BIPM, IEEE, et al) where appropriate prevents the ugly ambiguity. It only only seems to annoy people who already understand the issue at hand and dislike the prefixes for superficial reasons or are totally, utterly confused themselves. For further arguments made by myself and others at various times in the past, see this talk page archive. (specifically, here, here, here, and here (note the list at the end)) P.S. - Planetary Chaos' comments seem nearly comical considering the argument he (seems) to be attempting to make... Come on, folks, at the very least read the related Wikipedia articles on this matter before trying to express your opinion. P.S.2 - Shame on the folks that patronized Sarenne by labeling his obviously good-faith edits as vandalism and engaging in revert warring. -- mattb @ 2007-01-23T00:17Z
Yes, and K as in Kilo has been used to mean 1024, not 1000, in computer contexts, long before the introduction of the PC, too. See, for example, the IBM 1130 System Summary, which says on page 1 (PDF page 5) that "Core storage capacity is designated by a letter following the 1131 model number; A = 4k, B = 8k, C = 16k, and D = 32k.", and on page 2 (PDF page 6), speaks of machines with 4096, 8192, 16384, or 32768 words of memory, and, for machine that actually have megabytes of memory, see the IBM System/370 System Summary, which, on page 6-15 (PDF page 78), speaks of machines with "real-storage capacities of 1,024K (1,048,576 bytes) to 8,092K (8,388,608 bytes)". Guy Harris 07:21, 23 January 2007 (UTC)
Interestingly, for 2^10 bytes, the abbreviation "KB" is most common, which is arguably not using an SI prefix at all (SI defines a lower-case 'k' to mean 10^3). -- mattb @ 2007-01-23T14:11Z

Who here has bought a computer that uses the terms GiG or MiB? Any one? Any one at all? I just bought this computer two weeks ago..a 2007 PC and nowhere does it say GiG or MiB. On another note MiB might be confused with (Men In Black).  Planetary Chaos  Talk to me  00:33, 23 January 2007 (UTC)

Have you bothered to read half the links that have been provided, or are you more concerned with defending your original viewpoint than trying to understand why many people here are disagreeing? Let me say this explicitly: you yourself have confused terms and arithmetic several times here, so it's more than slightly ironic for you to counsel others to do their homework. -- mattb @ 2007-01-23T00:46Z
(in reply to Planetary Chaos) I'm not really considering writing any article on any of my computers, but even if I were telling you the capacities of the system, I would be sure to use the correct prefixes. For example, this laptop has 512 MiB of memory and an 80 GB HD. If I told you it had 512 MB of memory, I would be wrong. Not far wrong, granted, but wrong. "Mega" is a prefix that far predates computing, and it means one million of whatever follows it. Granted, the XiB terms are newer-that's to be expected, we only now have need of them, as only now are numbers such as 2^10, 2^20, and so on, important on a wide scale. However, they are clearly defined, verifiable, international standards, approved by multiple standards bodies, which have a clear, verifiable, and unambiguous meaning. Further, we can make trivial and obviously-correct changes to source wording, as commonly accepted when paraphrasing. For example, some companies may refer to their laptops as "laptop" computers, others to "portable" computers. If we decided that, for stylistic reasons, we should use only the term "laptop", that would be a perfectly acceptable paraphrase. Similarly, let's say an American source refers to an "organization" in an article which already uses British spelling. In that case, it would be perfectly acceptable to change that to "organisation"-we have not changed the meaning of what was said in our paraphrase. In the same vein here, when Dell refers to MB in terms of things that are always measured in MiB, we can safely presume that they mean MiB, and paraphrase for stylistic reasons and to avoid confusion. Dell may wish to speak in "layman's terms" when writing promotional material, and for that purpose, saying "1 GB of RAM" is certainly "close enough". However, in terms of writing an encyclopedia, we should work to be more precise, and to use exact numbers and terminology. Seraphimblade 02:54, 23 January 2007 (UTC)
Words take on different meanings in different contexts. In computing, the long-established meaning of the prefix "kilo" is 1024. Recently, there has been an attempt to re-define the prefix to accord with the non-computing meaning of 1000, and to invent a new prefix, "kibi", to replace what was conventionally denoted "kilo". This attempt is so far only partially successful, so it makes no sense to speak of "correct" or "incorrect", any more than it makes sense to speak of either or both "organisation" and "organization" as "incorrect spelling". The argument is between using something unconventional within computing, or something unconventional outside of computing, when writing about computers, and both approaches have their pros and cons. --Delirium 10:33, 23 January 2007 (UTC)
Quite so. The labels "correct" and "incorrect" are overblown here. My point (and others') every time this comes up is that strictly adhering to SI definitions and using IEC binary prefixes clears up any possible ambiguity assuming the writer is competent. I feel that the advantages of removing this ugly ambiguity outweigh any minor confusion that can be alieviated by reading the article that should be linked with the first usage of binary byte prefixes. I think it's far more advantageous for a reader to notice an unfamiliar abbreviation like MiB, click it, and learn about this notation issue than to see (excuse my own opinion here) misused SI prefixes and go on blissfully unaware that there is a difference between 10^6 and 2^20. If it came to a vote again, my position is clear (though I don't think it should have to since this has been discussed many times now). -- mattb @ 2007-01-23T13:27Z
  • One MegaMinute! I think we're writing a professional encyclopedia here and we don't need to use incorrect definitions if we know how to use the correct versions. I don't expect you would open a copy of Encyclopedia Britanica and find that different articles on the same subject would be using different definitions for the same units. Switching Imperial and Metric units is a poor decision as well. I think the entire encyclopedia should be written in International standard units. I am pushing for a wikimedia software feature to allow a user to set a units preference in this feature request. 13:47, 23 January 2007 (UTC)
Again, the label "correct" is overblown. While I tend to agree with you that we should only use SI prefixes with their strict SI definitions as set by the BIPM, there's the equally valid point that these SI prefixes have become nearly ubiquitous in common usage with computers (even professionals and engineers who know the issue often use them). It doesn't get us anywhere to insist upon which is the more correct usage, but we should consider which is the most accurate and uniform, and ambiguity plays a role in that. -- mattb @ 2007-01-23T14:09Z
Seraphimblade, Take alook at my second reply on this discussion.It kind of says what you said.
Mat, what is uniform? If you are speaking of MiB being more uniform then MB. I think you are wrong,How ever, MiB for something like 512 would be more accurate then MB.But on the other hand...MB would be more accurate for 500 then that of MiB.  Planetary Chaos  Talk to me  17:12, 24 January 2007 (UTC)
I'm really not sure what your question is... Can you express yourself with more lucidity? I fail to see much significant about the numbers you gave. The issue at stake is that using the prefix 'M', for example, in both the SI (10^6) and binary (2^30) contexts throughout Wikipedia is inconsistant. That's all I meant by "uniformity". I don't believe that it's good style to assume that the reader will perceive the correct context every time, or indeed even realize that there IS a context issue. Just as Wikipedia uses SI units in scientific articles for uniformity and accuracy, usage of the IEC binary prefixes where appropriate brings the same type of standard to computer and digital communications articles. (speaking of digital communications, just think about all the confusion there is between a kilobit, a kilobyte, a kibibyte, and a kibibit... Broadband companies make their living off this) -- mattb @ 2007-01-24T18:23Z
MiB for multiples of 1024*1024 (1048576) would be more accurate that MB. MB for multiples of 1000*1000 (1000000) would be more accurate than MiB. 500*1024*1024 (524288000) is 500 MiB, not 500 MB, just as 512*1000*1000 (512000000) is 512 MB. It's the units, not the multiple, that matters. Perhaps 500 is more commonly a multiple of 1000000 than of 1048576, and 512 is more commonly a multiple of 1048576 than of 1000000, but it's still the units that matter. Guy Harris 07:40, 25 January 2007 (UTC)
In the real world of talking about personal computer tech specs, nobody, nil, no one, not one single person refers to xiB units. Wikipedia articles, as they are written for non-experts, shouldn't be any different. I haven't been able to find even one other encyclopedia, magazine, or manufacturer website that refers to "gibibytes" of hard disk space or "mebibytes" of RAM. As xB units are the accepted norm, at least for now, changing to xiB units would only result in reader confusion. Sjenkins7000 04:27, 26 January 2007 (UTC)
You should take a look at a PC under Linux.Sarenne 13:29, 26 January 2007 (UTC)
If we used PC manufacturer marketing drivel as a benchmark, I can't even imagine how poor our standards of quality would be. -- mattb @ 2007-01-26T04:45Z

Should we have a vote on this? I think that everything was said. Current manual of style specifies very well where binary prefixes are to be used and still some editors are opposing it saying there isn't consensus. The only valid point why not to use them is IMO that it's WP:OR but there is no disscusion about accuracy. So we should decide whether we are about to use correct values or ones that are used in marketing materials of the computer industry (and may be utilize more info from computer marketing then since it's WP:OR not to call Apple' computers the best ones when Apple does so;). I don't think new consensus can be achieved by discussion, it has been discussed in the past (and not once) and nothing new has been said here. So we should probably vote to get over with this.--Pethr 11:29, 26 January 2007 (UTC)

This discussion titled "Converting to SI prefixes" continues on Archive 66.

go to Archive 66 >>