Bigger number space

[martinheidegger]

Thinking about how to increase the space to reduce collisions, see #23 (comment), I have been wondering if it wouldn't be possible to utilize two (a bit more out-there) concepts to increase the number space available and viability a little.

For one, as far I can see in rfc4144#4.1.3 the MSB 0 is not yet used and in this draft adds the version 0b1000. What if instead the language was changed to use MSB 0 as indicator for this version and MSB 1 ~ 2 for the var add MSB 3 to the random block and do the same for the 2 var bits.

    0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                           unix_ts_ms                          |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |          unix_ts_ms           | ver |         rand_a          |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                            rand_b                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                            rand_c                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

This would increase the random space to 77 bits, not quite on the 80 bit level of ULID but it should still be significantly more practicable?!

[edo1]

Do you mean

    0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                           unix_ts_ms                          |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |          unix_ts_ms           |1 x x x|       rand_a          |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |1 0 x|                      rand_b                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                            rand_b                             |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

where x is used as a random/counter?

P. S. There was a similar idea to use 0b111 variant with no version filed: #26

[martinheidegger]

Thank you, it's indeed a duplicate of the conversation