> I believe that the main purpose of documentation, help etc is
> to get the
> information across in a way that is most easily understood, ie that
> minimises the number of support questions.. ..and everyone
> surely knows
> what GB, MB and KB stand for. So let's leave it at that.
> Where's the "i"
> in "megabyte" ? Or is 1MiB 1000000 bytes, rather than 1048576?
1 MB isn't 1048576.
it's 1000000
mega isn't 2^10, it's 10^6
so where are YOU coming from?
(no, i'm not arguin, i don't particularly care. but i'm
pointing out that some people have completely firmly set
definitions and some other people also have firm definitions
and neither will agree the other's right. MiB is the international
standard for a 2^10 B(yte) specification. so if you mean
2^10 bytes, you mean MiB, not MB, even if you don't like it :)
So these are very very good changes :)
Dana Lacoste
Ottawa, Canada (a metric country :)
On Thursday 20 December 2001 10:36 am, Dana Lacoste wrote:
> > I believe that the main purpose of documentation, help etc is
> > to get the
> > information across in a way that is most easily understood, ie that
> > minimises the number of support questions.. ..and everyone
> > surely knows
> > what GB, MB and KB stand for. So let's leave it at that.
> > Where's the "i"
> > in "megabyte" ? Or is 1MiB 1000000 bytes, rather than 1048576?
>
> 1 MB isn't 1048576.
>
> it's 1000000
>
> mega isn't 2^10, it's 10^6
>
> so where are YOU coming from?
>
> (no, i'm not arguin, i don't particularly care. but i'm
> pointing out that some people have completely firmly set
> definitions and some other people also have firm definitions
> and neither will agree the other's right. MiB is the international
> standard for a 2^10 B(yte) specification. so if you mean
> 2^10 bytes, you mean MiB, not MB, even if you don't like it :)
This "international" standard seems to have excluded a few countries.
It wasn't until it was SET that I even heard of its existance. (And
then only through SLASHDOT!)
Everyone I know has been using KB/MB/GB for 1024 forever. The *only*
exception is networking, and the occasional FLASH/ROM size. The latter
isn't very common discussion, and among those that it is, they'd know
what the other was talking about. For the former, I can distinguish
easily depending on who it is.
Someone without a lot of experience: I have a 1MB connection. (this
user has a 1 Megabit connection)
Someone with experience: I have a 1mb/Mb connection. (This person has a
1 megabit connection has used a "standard" abbreviation.)
Know how these standards came about?
Actual use. Not a bunch of "engineers" in a room arguing over how best
to cause absurd changes in kernel help files.
>
> On Thursday 20 December 2001 10:36 am, Dana Lacoste wrote:
> > > I believe that the main purpose of documentation, help etc is
> > > to get the
> > > information across in a way that is most easily understood, ie that
> > > minimises the number of support questions.. ..and everyone
> > > surely knows
> > > what GB, MB and KB stand for. So let's leave it at that.
> > > Where's the "i"
> > > in "megabyte" ? Or is 1MiB 1000000 bytes, rather than 1048576?
> >
> > 1 MB isn't 1048576.
> >
> > it's 1000000
> >
> > mega isn't 2^10, it's 10^6
> >
> > so where are YOU coming from?
> >
> > (no, i'm not arguin, i don't particularly care. but i'm
> > pointing out that some people have completely firmly set
> > definitions and some other people also have firm definitions
> > and neither will agree the other's right. MiB is the international
> > standard for a 2^10 B(yte) specification. so if you mean
> > 2^10 bytes, you mean MiB, not MB, even if you don't like it :)
>
> This "international" standard seems to have excluded a few countries.
> It wasn't until it was SET that I even heard of its existance. (And
> then only through SLASHDOT!)
Yeah, no shit? The first time I buy 512MB of RAM and get 512000 KB
(aka 512000000 bytes) I am gonna be *PISSED*
/Mike
On Thu, Dec 20, 2001 at 11:13:52AM -0800, Nicholas Knight wrote:
> On Thursday 20 December 2001 10:36 am, Dana Lacoste wrote:
> > > I believe that the main purpose of documentation, help etc is
> > > to get the
> > > information across in a way that is most easily understood, ie that
> > > minimises the number of support questions.. ..and everyone
> > > surely knows
> > > what GB, MB and KB stand for. So let's leave it at that.
> > > Where's the "i"
> > > in "megabyte" ? Or is 1MiB 1000000 bytes, rather than 1048576?
> >
> > 1 MB isn't 1048576.
> >
> > it's 1000000
> >
> > mega isn't 2^10, it's 10^6
> >
> > so where are YOU coming from?
> >
> > (no, i'm not arguin, i don't particularly care. but i'm
> > pointing out that some people have completely firmly set
> > definitions and some other people also have firm definitions
> > and neither will agree the other's right. MiB is the international
> > standard for a 2^10 B(yte) specification. so if you mean
> > 2^10 bytes, you mean MiB, not MB, even if you don't like it :)
>
> This "international" standard seems to have excluded a few countries.
> It wasn't until it was SET that I even heard of its existance. (And
> then only through SLASHDOT!)
>
> Everyone I know has been using KB/MB/GB for 1024 forever. The *only*
> exception is networking, and the occasional FLASH/ROM size. The latter
> isn't very common discussion, and among those that it is, they'd know
> what the other was talking about. For the former, I can distinguish
> easily depending on who it is.
>
> Someone without a lot of experience: I have a 1MB connection. (this
> user has a 1 Megabit connection)
>
> Someone with experience: I have a 1mb/Mb connection. (This person has a
> 1 megabit connection has used a "standard" abbreviation.)
You have a 1 millibit per Megabit connection?!?!
[snip]
/David Weinehall
_ _
// David Weinehall <[email protected]> /> Northern lights wander \\
// Maintainer of the v2.0 kernel // Dance across the winter sky //
\> http://www.acc.umu.se/~tao/ </ Full colour fire </
[email protected] wrote:
> Someone without a lot of experience: I have a 1MB connection. (this
> user has a 1 Megabit connection)
>
> Someone with experience: I have a 1mb/Mb connection. (This person has a
> 1 megabit connection has used a "standard" abbreviation.)
Actually a 1 Mb/s connection is 1024000 bits/second (ie not 1000000 or
1048576 bits/second).
This came about because a basic voice channel is 64kb/s = 64000
bits/second. These are aggregated up into 32 channels at a time which
is known as an E1 in Europe. An E1 is known as a 2 "meg" connection
though it is 2048000 bits/s. Perhaps the correct appellation is 2
kkib/s?
Poor confused telecoms engineers ;-)
My personal view is that the kiB MiB GiB etc are very ugly but we
should grin and bear it to banish this decimal binary confusion
forever. Either that or put a note somewhere saying that all k, M and
G are 2^10, 2^20 and 2^30 and forget about it for a generation...
--
Nick Craig-Wood
[email protected]
[snip]
David Weinehall wrote:
>>>(no, i'm not arguin, i don't particularly care. but i'm
>>>pointing out that some people have completely firmly set
>>>definitions and some other people also have firm definitions
>>>and neither will agree the other's right. MiB is the international
>>>standard for a 2^10 B(yte) specification. so if you mean
>>>2^10 bytes, you mean MiB, not MB, even if you don't like it :)
>>>
>>This "international" standard seems to have excluded a few countries.
>>It wasn't until it was SET that I even heard of its existance. (And
>>then only through SLASHDOT!)
>>
>>Everyone I know has been using KB/MB/GB for 1024 forever. The *only*
>>exception is networking, and the occasional FLASH/ROM size. The latter
>>isn't very common discussion, and among those that it is, they'd know
>>what the other was talking about. For the former, I can distinguish
>>easily depending on who it is.
>>
>>Someone without a lot of experience: I have a 1MB connection. (this
>>user has a 1 Megabit connection)
>>
>>Someone with experience: I have a 1mb/Mb connection. (This person has a
>>1 megabit connection has used a "standard" abbreviation.)
>>
>
> You have a 1 millibit per Megabit connection?!?!
Yeah, she/he uses compression! *ROFL* Sorry, couldn't resist! :)))
bye,
Wilfried
--
Terorists crashed an airplane into the server room, have to remove
/bin/laden. (rm -rf /bin/laden)
On Thursday 20 December 2001 03:12 pm, Wilfried Weissmann wrote:
> [snip]
>
> David Weinehall wrote:
> >>>(no, i'm not arguin, i don't particularly care. but i'm
> >>>pointing out that some people have completely firmly set
> >>>definitions and some other people also have firm definitions
> >>>and neither will agree the other's right. MiB is the
> >>> international standard for a 2^10 B(yte) specification. so if
> >>> you mean 2^10 bytes, you mean MiB, not MB, even if you don't like
> >>> it :)
> >>
> >>This "international" standard seems to have excluded a few
> >> countries. It wasn't until it was SET that I even heard of its
> >> existance. (And then only through SLASHDOT!)
> >>
> >>Everyone I know has been using KB/MB/GB for 1024 forever. The
> >> *only* exception is networking, and the occasional FLASH/ROM size.
> >> The latter isn't very common discussion, and among those that it
> >> is, they'd know what the other was talking about. For the former,
> >> I can distinguish easily depending on who it is.
> >>
> >>Someone without a lot of experience: I have a 1MB connection. (this
> >>user has a 1 Megabit connection)
> >>
> >>Someone with experience: I have a 1mb/Mb connection. (This person
> >> has a 1 megabit connection has used a "standard" abbreviation.)
> >
> > You have a 1 millibit per Megabit connection?!?!
>
> Yeah, she/he uses compression! *ROFL* Sorry, couldn't resist! :)))
engineer 1: how can we drive network users more insane?
engineer 2: new compression scheme!
engineer 1: yeah! how about we add a bunch of random data to the end of
every packet?
engineer2: brilliant!
3 months later:
Nicholas Knight introduces the world to mPM compression.
(translation: DOH!)
> Yeah, no shit? The first time I buy 512MB of RAM and get 512000 KB
> (aka 512000000 bytes) I am gonna be *PISSED*
Have a work with your hard disk manufacturer then
>
> > Yeah, no shit? The first time I buy 512MB of RAM and get 512000 KB
> > (aka 512000000 bytes) I am gonna be *PISSED*
>
> Have a work with your hard disk manufacturer then
>
That isn't quite so important. My kernel isn't likely to f*ck up when
a 40GB HD = 40,000,000,000. I'm sure it will die quite painfully with
RAM chips that are not powers of 2.
Regards,
/Mike
On Fri, Dec 21, 2001 at 12:50:55PM -0500, Mike Harrold wrote:
> That isn't quite so important. My kernel isn't likely to f*ck up when
> a 40GB HD = 40,000,000,000. I'm sure it will die quite painfully with
> RAM chips that are not powers of 2.
Hell, your kernel isn't even going to barf if the "40GB" disk turns
out to be 39,501,824, or some other less than 40GB-of-any-flavor
value. Why do a version of "40GB" that means 40,000,000,000 when
disks are *never* that size anyway?
Just because disk manufacturers are, um, creatve, with their marketing
numbers, do we have to mess with the numbers that are trustworthy?
-kb, the Kent who is not so sure he has *ever* seen anything in a
computer that really was such a big round decimal number, but the Kent
who sees precise round binary numbers frequently.
>
> Hell, your kernel isn't even going to barf if the "40GB" disk turns
> out to be 39,501,824, or some other less than 40GB-of-any-flavor
> value. Why do a version of "40GB" that means 40,000,000,000 when
> disks are *never* that size anyway?
>
If you would pay more attention, you can see that on most drives there is
a small note that says: 1MB = 1000000 bytes. This is why the drive
capacity is smaller than the manufacturer says.
> Just because disk manufacturers are, um, creatve, with their marketing
> numbers, do we have to mess with the numbers that are trustworthy?
>
>
> -kb, the Kent who is not so sure he has *ever* seen anything in a
> computer that really was such a big round decimal number, but the Kent
> who sees precise round binary numbers frequently.
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
On Fri, Dec 21, 2001 at 08:49:44PM +0200, [email protected] wrote:
> If you would pay more attention, you can see that on most drives there is
> a small note that says: 1MB = 1000000 bytes. This is why the drive
> capacity is smaller than the manufacturer says.
So you are saying that my "12GB" drive is 12,000,000,000 bytes instead
of 12,884,901,888 bytes?
The drive in the notebook I am typing on now seems to be neither. If
I am doing arithmetic and reading hdparm output right, I think it is
12,072,517,632 bytes (smaller once formatted). Not a very round
decimal number.
My point was that big round decimal numbers are rare in computers, so
why do we suddenly need big round decimal units for talking about
computers?
Disk drives have inherently binary capacities, the only reason to
quote their capacities in decimal was to make them look bigger. I
don't see why we should have new units to make that easier. This is
particularly ironic when disk manufacturers are so good at making them
bigger at a pace that has seriously out-paced Moore's Law.
-kb
Mike Harrold writes:
>>> Yeah, no shit? The first time I buy 512MB of RAM and get 512000 KB
>>> (aka 512000000 bytes) I am gonna be *PISSED*
>>
>> Have a work with your hard disk manufacturer then
>
> That isn't quite so important. My kernel isn't likely to f*ck up when
> a 40GB HD = 40,000,000,000. I'm sure it will die quite painfully with
> RAM chips that are not powers of 2.
You'd be buying 537 MB of RAM, not 512 MB of RAM. I expect that
we will see this soon, since a binary GB has a 7% error.
(For kB the error was only 2.4%, which didn't matter so much.)
I would be selling RAM this way. It's stupid to do otherwise.
Consumers will prefer the bigger numbers.
Prefixes need to align with our number system. Unfortunately we
don't use something sane like hex. We use decimal, which is as
bad as base-9 or base-14. Oh well. Historical reasons you know,
and computers aren't bit-wise addressable either. We live with
this brokenness and can't afford to fix it all. So we might as
well use a notation, the base-10 prefixes, that is consistent
with our cummy number system.
On Thu, 20 Dec 2001 [email protected] wrote:
> Actually a 1 Mb/s connection is 1024000 bits/second (ie not 1000000 or
> 1048576 bits/second).
But gigabit ethernet is clocked at 1.25GHz with 8b10-encoding, meaning
you'll get literally 1.000.000.000 bits/second over that line. As far as I
know this is true for all kinds of ethernet.
Basically, it's only when it comes to memory terms that we use 1024 as a
base.
--
Mikael Abrahamsson email: [email protected]
Also sprach Mikael Abrahamsson
>On Thu, 20 Dec 2001 [email protected] wrote:
>> Actually a 1 Mb/s connection is 1024000 bits/second (ie not 1000000
>> or 1048576 bits/second).
>But gigabit ethernet is clocked at 1.25GHz with 8b10-encoding, meaning
>you'll get literally 1.000.000.000 bits/second over that line. As far
>as I know this is true for all kinds of ethernet.
>Basically, it's only when it comes to memory terms that we use 1024 as
>a base.
Again..."Uhm...no."
Ethernet isn't the only networking technology out there.
--
Jeff McAdams Email: [email protected]
Head Network Administrator Voice: (502) 966-3848
IgLou Internet Services (800) 436-4456
On Sat, 22 Dec 2001, Jeff Mcadams wrote:
> Also sprach Mikael Abrahamsson
> >On Thu, 20 Dec 2001 [email protected] wrote:
> >> Actually a 1 Mb/s connection is 1024000 bits/second (ie not 1000000
> >> or 1048576 bits/second).
>
> >But gigabit ethernet is clocked at 1.25GHz with 8b10-encoding, meaning
> >you'll get literally 1.000.000.000 bits/second over that line. As far
> >as I know this is true for all kinds of ethernet.
>
> >Basically, it's only when it comes to memory terms that we use 1024 as
> >a base.
>
> Again..."Uhm...no."
>
> Ethernet isn't the only networking technology out there.
While there are, historically 1Mb/s has allways been 1.000.000. The
misconception about it not being 1.000.000 is cause people associate it
with bytes. Though it's not because bits make up bytes that bits are
naturally forced to "live" on byte boundaries. As clock pulse generators
generally don't really live on byte boundaries either there was never a
real reason to make 1Mb/s related to bytes (or to make 1Kb/s related to
bytes). When referring to byte-bound data transfer speed you can stick to
xB/s instead of xb/s.
As an example, your Serial port doesn't exactly think of bits in
groups of 1024 either.
So in general your best bet is to see 1Kb/s as 1.000 bits per second and
1Mb/s as 1000Kb/s or 1.000.000b/s. As most technologies will stick to
that. Though off course through the ages a lot of things have been altered
it and therefor have added to the confusion.
Dirk Moerenhout ///// System Administrator ///// Planet Internet NV
> with bytes. Though it's not because bits make up bytes that bits are
> naturally forced to "live" on byte boundaries. As clock pulse generators
> generally don't really live on byte boundaries either there was never a
> real reason to make 1Mb/s related to bytes (or to make 1Kb/s related to
> bytes). When referring to byte-bound data transfer speed you can stick to
> xB/s instead of xb/s.
It gets worse the deeper you go. Over an HDLC based link for example
sequences of five one bits take longer to send due to bitstuffing. Any
networking terminology is generally grossly simplified.
At 03:20 PM 12/22/01 +0000, Alan Cox wrote:
>It gets worse the deeper you go. Over an HDLC based link for example
>sequences of five one bits take longer to send due to bitstuffing. Any
>networking terminology is generally grossly simplified.
Interestingly, the subject of bit-stuffing came up during the discussion of
modem testing, because of the HDLC framing used in V.42 error
control. Statistics provided by the representative from Hayes (R.I.P.)
showed that bit-stuffing occurred in roughly 1 out of 64 bits in the data
path when V.42 bis data compression was enabled. The 1:63 ratio held over
a surprisingly wide range of data patterns, all the way from repeated
single characters through text in English, French, German, and Chinese to
the output of a 64-bit random number generator. When v.42 bis data
compression was disabled, the average ratio was about the same but there
was considerably more spread because of data pattern sensitivity.
Which is grossly off-topic but I thought a few of you might be interested.
Stephen Satchell
> So in general your best bet is to see 1Kb/s as 1.000 bits per second
and
> 1Mb/s as 1000Kb/s or 1.000.000b/s. As most technologies will stick to
> that. Though off course through the ages a lot of things have been
altered
> it and therefor have added to the confusion.
I'd rather think of 1 kpbs than 1 Kbps...
K is Kelvin, and nothing else (IIRC). K is no prefix.
My proposal: humans should start using sedecimal as
primary numbering system. (And forget about octal
as fast as possible - it is referred way too often in UNIX!)
Greetings from snowful Bonn (Rhein)
-mirabilos
On Sat, Dec 22, 2001 at 05:43:41PM -0000, mirabilos {Thorsten Glaser} wrote:
> > So in general your best bet is to see 1Kb/s as 1.000 bits per second
> and
> > 1Mb/s as 1000Kb/s or 1.000.000b/s. As most technologies will stick to
> > that. Though off course through the ages a lot of things have been
> altered
> > it and therefor have added to the confusion.
>
> I'd rather think of 1 kpbs than 1 Kbps...
> K is Kelvin, and nothing else (IIRC). K is no prefix.
Some time ago, k was 1000 and K was 1024, b was bits and B was bytes ...
but then came the mega and giga, and you can't uppercase those ...
>
> My proposal: humans should start using sedecimal as
> primary numbering system. (And forget about octal
> as fast as possible - it is referred way too often in UNIX!)
>
> Greetings from snowful Bonn (Rhein)
> -mirabilos
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
--
Vojtech Pavlik
SuSE Labs