You might be surprised. Your ear focuses overly heavily on vowels, and both “b” and “t” are short-enough sounds your brain will happily confuse the two–especially if you were already expecting to hear one of the two. It’s basically your brain doing error correction–and, like error correction, it sometimes fixes the error in the wrong way. There’s a reason why the NATO alphabet exists, why its letter names are so wildly different from each other, and even why some of the letters or numbers have such odd pronunciations (check out 4 (FOW-er), 3 (TREE), O (OSS-KAH), etc.).
That all said, I’ve never really grokked proquints. As the grandparent is pointing out, they don’t seem to have learned from any of the research that went into the NATO alphabet in the first place. I’m very confident I can read the numbers in the article with fewer errors if both I and the person I’m working with do indeed just stick to NATO phraseology.
Obligatory plug for the EFF word lists, one of which has 6**4 words that can be distinguished just by their first 3 letters: https://www.eff.org/deeplinks/2016/07/new-wordlists-random-passphrases – I use this all the time for generating keywords. And they’re all real words, no “haguz”.
Seems like the main advantage over word list approaches (e.g. correct battery horse staple) is that decoding them is simpler, but I’m not sure that matters to me.
I think I prefer word list approaches that use well-chosen words for easy speaking and writing.
We wonder why IP addresses aren’t encoded this way – they sure
would be easier to remember.
Sometimes the numerical values of IP addresses are relevant. For example, it seems to me that this representation would multiply the difficulty of understanding expressions in CIDR notation.
Dotted quads IP addresses, e.g. 127.0.0.1, will soon (someday?), with IPv6, be twice as long.
IPv4 addresses are 32 bits long. IPv6 addresses are 128 bits long. 128 = 32 × 4. As ASCII text, IPv4 addresses are 7 to 15 chars long and IPv6 addresses are at most 39 chars and can be shorter than IPv4 addresses. Maybe the author meant “twice as long” in text on average, but don’t see any evidence cited about average lengths of textual representations of IP addresses.
I once used mnemonicode, which “ converts 32 bits of data into 3 words from a vocabulary of 1626 words. The words have been chosen to be easy to understand over the phone and recognizable internationally as much as possible.”
Lower density, but a lot easier to transmit by voice. (I was using it for pairing, to allow two people to verify that they’ve exchanged public keys with no MITM.)
“Was that
budovortudov?”“
budovwith a ‘B’”“‘B’ as in ‘Boy’ or ‘T’ as in ‘Tom’?”
“‘B’ as in ‘Boy’”
“This is so much better than numbers.”
“I agree!”
What accent makes “b” and “t” sound the same?
You might be surprised. Your ear focuses overly heavily on vowels, and both “b” and “t” are short-enough sounds your brain will happily confuse the two–especially if you were already expecting to hear one of the two. It’s basically your brain doing error correction–and, like error correction, it sometimes fixes the error in the wrong way. There’s a reason why the NATO alphabet exists, why its letter names are so wildly different from each other, and even why some of the letters or numbers have such odd pronunciations (check out 4 (FOW-er), 3 (TREE), O (OSS-KAH), etc.).
That all said, I’ve never really grokked proquints. As the grandparent is pointing out, they don’t seem to have learned from any of the research that went into the NATO alphabet in the first place. I’m very confident I can read the numbers in the article with fewer errors if both I and the person I’m working with do indeed just stick to NATO phraseology.
Over less-than ideal audio channels—especially if there is no video component—they are easily confused.
Obligatory plug for the EFF word lists, one of which has 6**4 words that can be distinguished just by their first 3 letters: https://www.eff.org/deeplinks/2016/07/new-wordlists-random-passphrases – I use this all the time for generating keywords. And they’re all real words, no “haguz”.
That’s neat!
Amusingly, all of the examples look like they could be names of Dune characters!
Here are the example translations from the article, because they bury the lede.
Seems like the main advantage over word list approaches (e.g. correct battery horse staple) is that decoding them is simpler, but I’m not sure that matters to me.
I think I prefer word list approaches that use well-chosen words for easy speaking and writing.
Sometimes the numerical values of IP addresses are relevant. For example, it seems to me that this representation would multiply the difficulty of understanding expressions in CIDR notation.
IPv4 addresses are 32 bits long. IPv6 addresses are 128 bits long. 128 = 32 × 4. As ASCII text, IPv4 addresses are 7 to 15 chars long and IPv6 addresses are at most 39 chars and can be shorter than IPv4 addresses. Maybe the author meant “twice as long” in text on average, but don’t see any evidence cited about average lengths of textual representations of IP addresses.
Why is “r” a vowel?
I’ve just started writing names of identifiers in Toki Pona.
Creator’s Github repo with code and links to other projects is here: https://github.com/dsw/proquint
Urbit uses a similar model, and I find it completely unmemorable. [But maybe proquints are a bit better.]
I saw a system that uses identifiers like “malevolent-short-old-wolf”. Those are memorable.
(Also, it is a weird writing style to put the examples of proquints as the very last item in the article.)
I once used mnemonicode, which “ converts 32 bits of data into 3 words from a vocabulary of 1626 words. The words have been chosen to be easy to understand over the phone and recognizable internationally as much as possible.”
Lower density, but a lot easier to transmit by voice. (I was using it for pairing, to allow two people to verify that they’ve exchanged public keys with no MITM.)