A lot of use cases may not want the default behavior to be that labels are indexed. While I agree it should be an optional feature, in a graph where a lot of the nodes have the same label name you may not want all of the labels cluttering up your index. My vote would be for this feature, if added, to be off by default. Imagine you have a large social graph. A lot of the nodes in the graph will be labelled "person" - it is not always the case that you want all of those labels in the index. It is sometimes better to index and key off a more granular node property.