(No version information available, might only be in Git)
A unique identifier created for database objects. If an object is inserted into the database without an _id field, an _id field will be added to it with a MongoId instance as its value. If the data has a naturally occuring unique field (say, a username or timestamp) it is fine to use this as the _id field instead, and it will not be replaced with a MongoId.
Instances of the MongoId class fulfill the role that autoincrementing does in a relational database: to provide a unique key if the data does not natually have one. Autoincrementing does not work well with a sharded database, as it is impossible to find what the next number should be quickly. This class fulfills the constraints of quickly generating a value that is unique across shards.
Each MongoId is 12 bytes (making its string form 24 hexidecimal characters). The first four bytes are a timestamp, the next three are a hash of the client machine's hostname, the next two are the two least significant bytes of the process id running the script, and the last three bytes are an incrementing value.
MongoIds are serializable/unserializable. Their serialized form is similar to their string form:
MongoDB core docs on » ids.