More and more often, we see schema designs that use UUIDs as primary keys. That's a valid choice if you're concerned about sharding and partitioning your database, but it has its own drawbacks, sometimes in unexpected places.
