[Edited to reflect comments below]
I propose as a convention we abuse terms for talking about length of time to talk about particular powers of two of epoch lengths; this will make it easier to think about time within the protocol.
Note: 1 epoch = 384 seconds = 6.4 minutes
| Unit |
Epoch count |
Actual length |
| Hour (or "ethereum hour" to disambiguate) |
2**3 = 8 |
~51 minutes |
| Day |
2**8 = 256 |
~27.3 hours |
| Week |
2**11 = 2048 |
~9.1 days |
| Month |
2**13 = 8192 |
~36.4 days |
| Year |
2**16 = 65536 |
~291 days |
Note that the fact that most of these are "slightly too long" is arguably helpful for a possible future scenario where we cut slot times from 12 to 8 seconds once it proves safe. If slot duration drops to 8 seconds, hour and year would have their epoch count doubled (and become ~68 minutes and ~388 days, respectively) and the other units would keep their current epoch count (ethereum day ~= 18.2 Earth hours, ethereum week ~= 6.1 Earth days, ethereum month ~= 24 Earth days).
[Edited to reflect comments below]
I propose as a convention we abuse terms for talking about length of time to talk about particular powers of two of epoch lengths; this will make it easier to think about time within the protocol.
Note: 1 epoch = 384 seconds = 6.4 minutes
2**3 = 82**8 = 2562**11 = 20482**13 = 81922**16 = 65536Note that the fact that most of these are "slightly too long" is arguably helpful for a possible future scenario where we cut slot times from 12 to 8 seconds once it proves safe. If slot duration drops to 8 seconds, hour and year would have their epoch count doubled (and become ~68 minutes and ~388 days, respectively) and the other units would keep their current epoch count (ethereum day ~= 18.2 Earth hours, ethereum week ~= 6.1 Earth days, ethereum month ~= 24 Earth days).