Why I use UUIDs for primary/foreign keys

[eluser]Dam1an[/eluser]
First up, I'm going to say I use auto increment ID's, not cause I don't know any better, but cause I prefer not to have either strings or UUIDs as a primary key
Lets say you've got any sort of system where you access 'stuff' by passing the ID in the URL, such as users/123, much nicer then users/3857ab365-3857ab365-3857ab365-3857ab365
(another alterntive in the above example would be using the email as the PK, definatly don't want that as a link)

Also, you say it's very unlikely to get a collision, and they're unique accross databases as well, surely there will be a clash eventually (and without exhausting all 16^32 possibilities)

And lastly, going back to your main point, people who write libraries are unlikely to cater for a small number of users that use UUIDs when they have a much, much larger portion of users using standard auto increment IDs