Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

The Gibit


I can already hear the groans, but I'd like to bring up the Gibit.

As you may know, a year ago the IEC published new prefixes which 
recognize that what used to be called, for example, 1 Kbit is really 
1024 bits. Some new terms are:

kibibit		Kibit
mebibit		Mibit
gibibit		Gibit
tebibit		Tibit

As you know, the difference between the conventional computer usage 
and the actual powers of 1000 grows with each step; e.g. 1 Tibit is 
about 10% larger than 1 Tbit. Sometimes these differences are 
critical. Take, for example, the two data rates that the new "10 Gb" 
Ethernet PAR is addressing. There is real potential for confusion 
here. Besides, since many of our standards end up in IEC, it seems to 
me a real possibility that IEC will some day want us to conform.

[By the way, IEC doesn't recognize the abbreviation "b" for bit; they 
just use the word "bit".]

And don't let the pronunciation bother you; I just pronounce "10 
Gibit/s" as if it were "10 Gibits". A lot easier than "10 Gigabits 
per second."

So I think we need to start thinking about migrating from Gb/s to 
Gibit/s within 802.



P.S. Here are some references:

(1) IEEE Standards supports the new definitions and thinks there are 
significant inconsistencies with the conventional usage. For example, 
did you know that a 1 GB disk drive has 1 million bytes, not 
1024*1024 Bytes? See:

(2) NIST provides an excellent summary and a reference to the IEEE 
Standards position:

(3) A few months ago, some of my comments on this topic 
<> were 
included in Jeffrey Harrow's web-based column "The Rapidly Changing 
Face of Computing." Part of my note became the title of the article 
"Een schijf van 10 tebibyte" in the Science section of the Dutch 
daily newspaper _Reformatorisch Dagblad_ (May 4, 1999). It's a good 
article <>, if you read