Major Description
Currently we use a transaction's write pointer as the way it is identified everywhere to the transaction service. This works in the case of a single stage transaction (where a single write pointer is used), but will not work if we want to support checkpointed transactions which may have multiple write pointers. For checkpointed transactions we will want to clearly distinguish between the current write pointer being used (which will change after a checkpoint) vs. how the transaction is identified to the system (which will remain the same).