Marcus,
I don't have the bandwidth right now to carry the impl forward to Draft 11 - this is one of the pain points of UBJSON right now, it is still a little bit in flux and I'm waiting for discussion to settle down before doing a big rev on the implementation.
I had planned (and may still plan) for Draft 11 to be the final 1.0 release of UBJSON - the only portion of the spec under hot debate right now is allowing strongly-typed containers to contain other strongly-typed containers (turtles all the way down!) -
https://github.com/thebuzzmedia/universal-binary-json/issues/43
As annoying as this moving target is, the movement is getting less and less year over year which was always my goal.
The website only (intentionally) represents the latest spec and I don't have a clearcut enumeration of each version of the spec -- the reason for this is that the changes between the different drafts were not always nicely incremental; there was some backstepping and mild pivoting that took place around some of the more pivotal releases (Draft 4, 8, 10, etc.) and I don't want the confusion of [s] meaning "COMPACT STRING" and then 1 year later [s] meaning "SEGMENTED BYTE SEQUENCE" to exist in the same universe (this is just an example, this isn't actually the meaning of [s] now)
The benefit is that everyone new to the spec only learns the latest/most accurate representation.
The downside is what you recognized -- trying to actively work with previous drafts is hard (impossible?) but that is a cost I knowingly incurred for the spec... my goal is not to make working with previous drafts easier -- my goal is to focus first and foremost on the spec and understanding use cases across the scope of computing and then my followup goal to that is getting some really top tier implementations in all the major languages to enable it's use.
We are transitions from Step 1 (soldify spec) into Step 2, but I'd really like to wrap the decision up around strongly typed containers first because if we backstep there at all it will effect a lot of optimized scenarios, implementations and binary data support.
Hope that helps provide some context as to why this isn't a simpler answer. I do appreciate the interest in UBJSON; the folks that have leveraged it have gotten a lot of gains out of it.
Best wishes,