<id>

I think <id> should be a server-unique identifier.

What are the reasons for pathing?

1. To reference reusable ranges for enums.  There are other solutions such as range id or the reference could simply be the <id> of range's object.

2. To write to non-unique id's.  See, non-unique id's introduce problems!  The current way write is documented already has to be changed for pathing, it just doesn't work.  We should just eliminate nested writes, there is no benefit over batch writes.

What else?  What am I missing?