The PGP biometric word list is pretty interesting too:
I especially like how they have different words for odd and even positions in the number stream to act as a simple error-checking mechanism.
Can I use this for IPv6 addresses?
Actually the default dictionary has a problem. It’s sorted. This works well enough for uniformly distributed random data, but it’s a problem for using it for sequential data. A one bit difference will result in BAND becoming BANG. Ok, so maybe that’s a feature. All your IP address will have the same prefix. Then again, if you do it as /64, the net name will be fixed, with each host having a rather distinct human friendly name.
If I’m reading these into a phone (clearly not the intent) I want HO and HOE to be very far apart so as to not end up with almost correct but subtly different data.
this (making complex keys recognizable) is a hard problem. i spent a long time trying to make a graphical representation of keys. it’s hard to find something that (1) is memorable (2) is attractive (3) can be measured as being effective. the last point is important - even with text like this, two words may be close enough that people confuse them. the exact use case matters - checking explicitly with a second party is probably easier than displaying a site’s key so that a casual user notices “something has changed” one day.
ps the biggest lesson from (1) is that symmetry in images is important. all the synthetic icons (eg the auto-gen github icons) are symmetric. it’s way easier to remember and looks better.