It would be very helpful for clients, like replication or sharding, to track collection via a unique id which doesn't change during a rename or restarts. The "id" should be the same across replicas in a set or shards in a sharded cluster. Note: There is already an _id in sharding metadata for a collection but that is not exposed.

Here is an example of output from listCollections with an "id":

> db.runCommand("listCollections")
{
	"collections" : [
		{
			"id" : NumberLong("70368744177664"),
			"name" : "system.indexes",
			"options" : {}
		}
	],
	"ok" : 1
}