Whenever I run across the "byte order" debate, I'm reminded of the practice of "facing" money. When counting money, a bill may "face" one of four ways. Now many people can count money that is facing different ways, but most professional counters make sure the bills are all one way. It seems to make counting and recounting just a bit easier (and more accurate) if the bills are "faced."

The use of binary numbers and the creation of "base 2" binomials presents much the same problem. At its simplest, a two-place binomial may represent four "states," designated "00," "01," "10," and "11." If the states to be represented are sequential, we must decide whether the first state is designated 00 or 11. Additionally, we must decide whether the second state is created working from the right or left. The choices are 00-01-10-11, 00-10-01-11, 11-10-01-00, and 11-01-10-00.

All of this is compounded when we group the bits into an 8-bit "byte" which may represent 256 states (2^8.) The terms used for the two major methods of assigning "byte order" are "big endian" and "little endian" (from Jonathon Swift's "Gulliver's Travels.") Interesting discussions about the problem are David K. Every's What Is Endian?, James M. Curran's Little Endian vs. Big Endian, and William Stallings' Endian Issues.

The obvious choice would be to choose among equals and have everyone agree on one. Suffice to say that the computer industry has yet to resolve these issues and, in my opinion, the cost in performance is high.

Favorites: Worldwide Institute of Software Architects * about.com * howstuffworks.com * Janis Ian * GPS Tracker * New Yorker cartoon * David Coursey/Time * More ...