Re: 9.4 Proposal: Initdb creates a single table

David Fetter <david(at)fetter(dot)org> writes:
> On Thu, Apr 24, 2014 at 11:30:15AM -0400, Tom Lane wrote:
>> Essentially, that would mean carrying around our own implementation
>> of libuuid --- which includes a bunch of not-terribly-portable
>> stuff, such as discovering the machine's MAC address(es). That's
>> not really something I want to see us putting project manpower into.

> We don't need to do the not-terribly-portable stuff in the first
> round. For that, there could still be a bundled extension.

> The point is that UUIDs are nowhere near as usable as users have the
> right to expect, and we should fix that.

The reason that UUIDs aren't as usable as users "have a right to expect"
is that the underlying technology doesn't meet their (your) expectations.
Just because it's easy to imagine that there are universally unique
identifiers doesn't mean that there actually *are* universally unique
identifiers. There are only approximations with varying failure modes.

This is not our fault, and I don't want us to get caught up in trying
to fix a fundamentally broken concept --- which is what a generic
"uuidserial" API would be. If you try to paper over the difficulties
here, they'll just bite you on the rear someday.

regards, tom lane