On Wed, Dec 29, 2021 at 3:46 PM Tim Bray <tbray@xxxxxxxxxxxxxx> wrote:
I dunno, I'm in the Twitter @bluesky conversation, and the more I see the alternative proposals for scalable global-human-population identity schemes, the more I think telephone numbers aren't a terrible model. Adequately memorable, easy to scribble on a cocktail napkin, satisfactory as a unique key in your own personal contacts application whatever it may be, and sufficiently boring as to not provoke a land-grab rush. Yes, some numbers are cooler than others, but it's less intense an issue than having to be timb4857292@ rather than timbray@ because you arrived later on the scene.
If we were to redo telephone numbers, we would not do exchanges or area codes. They are no longer necessary and to the extent they still have meaning, it is privacy violating meaning.
We would probably use Base32 instead of just numbers. That gets us to some fairly impressive compact identifiers. 5 bits per character, an 8 character sequence is sufficient for a trillion identifiers.
Sure, from a technical point of view, @KXJL-M4AA is acceptable as an identifier. It is more usable than
@MB5S-R4AJ-3FBT-7NHO-T26Z-2E6Y which gives us 112 bits of work factor.
I still think that for Alice, @alice is going to be the ideal, @the_famous_alice a close approximation and so on. Sure that leads to more hassle for the registry than random assignment but the cost is much less memorable callsigns.
