I'm not a yang expert. The document is clear. I'm not sure how the dynamics of these allocations will play out over time, but it seems entirely reasonable to try this as planned.

One nit, the draft says:

   The following items are identified using SIDs:
   *  identities

That seems wrong, you're not intending that a 63-bit number be allocated for user@example.com so I think what you mean is that SIDs can identify types of identity/identifiers rather than the values of those. If the latter were the intent, then I'd have significant privacy concerns and this review would have reached a "not ready" conclusion.