The conditions for the use of source-generated UUIDs and receiver-generated UUIDs

[sergeyprokhorenko]

The conditions for the use of source-generated UUIDs and receiver-generated UUIDs

[kyzer-davis]

@sergeyprokhorenko, I see your point now about distributed node monotonicity and defining a way to ensure we have the best monotonic UUID in distributed node generation scenarios.

Coincidently I also received an email from the IETF mailer on the same topic of "Monotonicity guarantees with distributed nodes" which we don't define in the spec.

As such I have steered some others to this thread. Hopefully we have more feedback on this specific topic and can come an agreement on a solution for this scenario.

[kyzer-davis]

I am not 100% sure the proposed text provides anything more over the existing text.

• Integrity and feedback: I am not a fan of specifying that somebody generate a second UUID for every remote UUID since that is double the compute with little upside... or at least little upsides defined in the this text.
• Remote vs Local: IMO we already do a pretty good job of concisely stating the downsides local vs remote generation in the first line of the current text. The proposed text seems to split it into two sentences that more or less convey the same meaning as the original text.

[sergeyprokhorenko]

The proposed text is more detailed and gives more accurate instructions to database developers. It also adds the possibility to have two UUIDs if there are high requirements for monotony (this is not obligatory, but possible)