Not all Data-Centric is Web-Centric (probably the converse perhaps?)...   There's ton's of data crunching to be done well before anything ever spits it's way out onto the Web. At some point (way out in super abstract land perhaps) all of that reduces to graph traversal.  So why exactly are you going to have me learn Gremlin or Cypher or whatever when I already know xPath and XSLT which are more powerful than the other alternatives?