On Tue, 27 May 2025 21:44:32 GMT, Roger Riggs <rri...@openjdk.org> wrote:
>> spec updated in class javadoc to RFC 9562 which obsolete RFC 4122, @spec tag
>> also added above the method.
>
> I'd say "create" instead of "retrieve" in the first line comment. (Though
> that word is used in the other static factories).
>
> The "sub-millisecond precision" can't be relied upon. Its the precision that
> gives the impression that it can be added to the millisecond value and get an
> exact time. But the nano-second value is read from a different clock and the
> bits being used from it may wrap-around in the time between the reads of the
> clocks.
> The second description ("derived from") is makes fewer promises than the
> first-line comment.
> An application should not use them for any purpose other than random. And
> for that the buffer already contains random bytes.
>
> An alternative is to capture the MS time and the nano-time on first use to
> compute an offset and then use only the nano-time plus/minus the offset to
> create the version 7 UUIDs.
The `timestamp` method may mislead an app developer thinking its the time from
the version 7 timestamp.
The first-line comment might make it more obvious if it says:
`The timestamp value associated with a version 1 UUID.`
The Unix Epoch time would be easier to use if a method was added to return it.
`long UnixEpochTimeNanos()` (name subject to bike-shedding).
And throwing if it was not version 7.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/25303#discussion_r2110368888
