One of the potential roadblocks of multi-master replication is the fact that schemas must be compatible on all nodes. Any node that does not have the same columns and types on tables risks being unable to consume data from the upstream node that generated the incompatible tuple.
As a result, a hallmark feature of BDR (and likely any other multi-master system that emerges in the future) is that DDL is transparently replicated. This ensures that the schemas are always compatible with all nodes, and prevents costly delays in replication caused by stuck data.
This recipe will be a short proof of concept to prove that DDL is transparently replicated and serves as a good check to try on any new BDR cluster to ensure that it's working correctly.
