Details
Description
At the next major release, we should swap out the Java serialization work in VertexProgramHelper and instead, use the KryoShim work developed by dalaro. This means we need a KryoShimService in gremlin-core. I say we remove the HadoopPoolsKryoShimService and go with a GryoPoolKryoShimService that then Hadoop-based OLAP engines can use (as well as engines like TinkerGraph).
This would a a "minor breaking" change as I suspect no provider is that deep into the serialization API of TinkerPop and the change would be a simple "change X to Y" sort of change for them.