This question <https://groups.google.com/d/msg/gremlin-users/rPguKVm3GVE/f1OdIFvBAwAJ> on TinkerPop 3.3 on the Gremlin users list ultimately comes back to Scala 2.11 support <https://groups.google.com/d/msg/janusgraph-users/UypKyqnWqzc/eaQUOR2IDAAJ>. I can understand that some features slated for 3.3 might be far off ( TINKERPOP-1564 <https://issues.apache.org/jira/browse/TINKERPOP-1564>), but I don't think we should hold up on updating core dependencies such as Spark+Scala.
Azure Cosmos DB (based on TP 3.2.4) recently published <https://docs.microsoft.com/en-us/azure/cosmos-db/spark-connector-graph> about connecting it to Spark, and it tells the developer to build TinkerPop master (TP 3.3.0-SNAPSHOT) because Azure HDInsight uses Spark 2.0. We ran into a similar issue when TInkerPop 3.0 came out with the Hadoop 1 dependency rather than Hadoop 2, and I think a lot of users had already moved to Hadoop 2 <https://groups.google.com/d/msg/gremlin-users/_9UD6STUUOE/iEr2qoxFAQAJ> by then. Side note, I don't know what the schedule for Hadoop 3 is, but they already have alpha versions available. Spark 2.0 came out in July 2016, and it has been integrated on master since September 2016 <https://groups.google.com/d/msg/gremlin-users/j7lDGg5pIo8/QNWaFC3ABQAJ>. Spark 2 builds with Scala 2.11 by default. If we can't get the next TinkerPop release until next year, TinkerPop's Spark+Scala dependencies will be very far out of date. Would it make sense to cherry pick Spark 2.0 support into the tp32 branch?
