Ah, I see someone else has already brought up version 1 UUIDs. > What about invalid and reused MAC addresses and devices with no MAC address > at all?
Not an issue with SQLite since that part of the UUID is a constant within a given database. It would be reasonable to simply pick "0" then set the low bit of the first octet to 1 in accordance with the standard for non-MAC node IDs. The standard suggests using a hash or a cryptographically secure random number generator, setting the > What about time resets to the epoch which are not restored, user time changes, I know some systems at least increment the node each time a time change is detected. It will take 2^47 time changes to roll over. Since the node part is not relevant to SQLite, this is perfectly safe. > daylight saving or leap seconds? Not relevant to the timestamp format, since it's an epoch time. > It sounds even more probabilistic than the probabilistic methods. Does anyone > actually use it? It's Microsoft's standard GUID/UUID format and is also used by a lot of other systems for compatibility. _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users