Open jmalloc opened 4 years ago
Discussing some specifics of the persistence layer with @danilvpetrov, we determined that this couldn't proceed with the usage of the persistence system as it is.
The behavior of any transaction operation that uses OCC and thus returns some kind of ErrConflict
error would need to change to expect such an error at commit-time, rather than when the individual transaction operation is called.
Additionally, aggregate.Sink
relies on some un-guaranteed implementation specifics of the current persistence providers as detailed in #241.
This should be a lot easier now that DynamoDB supports transactions.