cg at cdegroot.com (Cees de Groot) wrote:
>  <goran.hultgren at bluefish.se> said:
> >You know - UUIDs is typically a thing you want to be able to trust...
> >It's sortof the whole point with them...
> Well, IMNSHO the UUID stuff that's currently in the image is
> broken. UUID's shouldn't be big random numbers, they should be *provable*
> universable unique. At the moment, the only *provable* universable
> unique scheme that is standardized is to have the machine's IP address,
> the process id, the current time and a sequence number in it.
> 
> Random UUID's are just a cheap cop-out for when you cannot (or are too lazy to
> find out or too shy to ask the user) get the IP address and/or the process ID.
> They should work, because the chance of re-generating the same UUID twice is
> small, but as you have experienced there is no guarantee. 
> 
> I think that UUID generation (more and more important in a distributed world)
> should be completed with IP based UUID's (where, that's the drawback, some
> human being needs to seed the thing with the IP address of the box).

Does anyone know how the primitive is implemented then? On Mac? John?

regards, Göran

PS. Perhaps I will just have to use something else.