carado.moe

tammy's blog about
AI alignment, utopia, anthropics, and more;

also on twitter, lesswrong, rss, and at

endiannesses

i extend the term "endianness" to mean the order of symbols within a representational sequence, and whether that order is least-significant to most-significant (little endian, hereby LE) or most-significant to least-significant (big endian, hereby BE).

in technology and culture, there are a variety of places in which endianness makes sense to talk about, and not all of them are of the same endianness.

• famously, and this is what let to the terms LE and BE being coined for numbers in the first place, bytes within a multiple-byte-wide integer in computers can have different endianness, typically depending on ISA; though LE has become the de-facto standard
• in computers, bits within a byte are LE; indeed, bit-shifting instructions address in order the least significant bit as 0, the second most significant bit as 1, the third as 2, and so on. this is why i believe LE is the correct choice for bytes within an integer: it's coherent with the order of bits within an integer.
• numbers in english are in BE: in the number "5005", the first "5" is more significant than the last "5". that said, english uses arabic numerals, which were still ordered with the most significant digit on the left even when they were used in arabic, which is read to right-to-left. this makes arabic numerals dependent on language, and originally LE, as it still in semitic languages.
• times of the day, such as "11:30" are in BE, though "5pm" is LE ("pm" is more significant than "5"), which makes "eleven thirty PM" mixed endian.
• dates in most western countries such as "31/12/2005" are LE, while dates in america such as "12/31/2005" are famously mixed endian; the recommended ISO-8601 dates are BE.
• postal addresses like "John Doe, 50th Central Avenue, Seattle, Washington, USA" are LE
• file paths such as "/home/carado/.bashrc" are BE, but domain names such as "something.carado.moe" are LE, making URLs (with a domain name followed by a file path) mixed endian.
• in the same vein as file paths, C-like programming language record access paths like customer.address.city, and indeed paths in almost all computering contexts i know of, are BE
• possessive sequences like the english "john's mother's cat" or the japanese "ジョンの母親の猫" are BE