Bengt Richter <b...@bokr.com> wrote:
> BTW, how did nix get such a weird alphabet for 0-31 ?
My guess is that the weird alphabet was chosen to avoid some of the most
common letters in English text, so that when scanning build outputs for
embedded hashes, one is less likely to mistake something else (e.g. text
or some other base32/base64 encoding) as a Nix hash. The omitted
letters in Nix hashes are (e t o u), whereas (e t a o) are the most
common letters in English text. I'm not sure why they chose to omit 'u'
though, given that it's quite far down the list of most common English
letters.
Regards,
Mark
