Re: BUG #5118: start-status-insert-fatal

"Kevin Grittner" <Kevin.Grittner@wicourts.gov> writes:
> Robert Haas <robertmhaas@gmail.com> wrote:
>> Well, then Tom's idea of using a random number seems pretty solid no
>> matter how you slice it.  Maybe a UUID.

> A random number is looking like the best option.  I'm not sure why I'd
> want to generate a perfectly good 128 bit random number and then throw
> away six of the bits to dress it up as a UUID, though.  Do the
> libraries for that do enough to introduce entropy to compensate for
> the lost bits?  Any other benefit I'm missing?

I was envisioning just using PostmasterRandom() (after initializing
the seed from time(NULL) as we do now).  I don't think we need a
super-wide random number.

            regards, tom lane