JéRôME CARRETERO WROTE:
>> SO YOU RECOMMEND ADDING A 'PARENT' ATTRIBUTE TO THE
<THREAD>
>> ELEMENT IN RFC3921BIS, CORRECT? I THINK THAT CHANGE
WOULD BE
>> BACKWARDS-COMPATIBLE, WE JUST NEED TO MAKE SURE
THAT IT IS CLEARLY
>> SPECIFIED.
>
> YES. I SEE THAT CHANGE HAS ALREADY BEEN MADE ;)
PROVISIONALLY, YES. FEEDBACK IS WELCOME.
> I'D ALSO SUGGEST THE USE OF TIME-BASED UUIDS, BECAUSE :
>
> - FOR AN
> INSTANT MESSAGING APPLICATION, THE RISK OF COLLISION IS
NULL SINCE 2
> DIFFERENT MACHINES COULDN'T GENERATE THE SAME UUID, AND
IT'S
> UNTHINKABLE THAT ONE MACHINE COULD START 2 THREADS IN A
FEW 10^-8S.
> THIS RISK IS PRESENT (ALBEIT RIDICULOUS) WITH
€œFULLY-RANDOM€ UUIDS.
>
> - TIME-BASED UUIDS COULD BE CONVENIENT FOR BEING ABLE
TO SORT CHILD
> THREADS BY DATE WITHOUT RELYING ON MORE INFORMATION IN
THE MESSAGES
> (YES, THIS IS A BIT CONTRARY TO THE OPAQUENESS OF THE
THREADID...).
>
> OF COURSE IT FORCES IMPLEMENTATIONS TO USE TIME-BASED
UUIDS, WHICH
> COSTS A FEW MORE LINE OF CODES IF DONE FROM SCRATCH...
AND MAY BE
> WISHFUL THINKING SINCE SOME IMPLEMENTATIONS MIGHT
ALREADY EXIST
> TODAY.
WHAT DO YOU MEAN BY "TIME-BASED UUID"? COULD YOU
SHOW US AN EXAMPLE OR
PROVIDE A REFERENCE TO A SPECIFICATION? I AM NOT FAMILIAR
WITH THIS CONCEPT.
THANKS!
PETER
--
PETER SAINT-ANDRE
HTTPS://STPETER.IM/
|
