What is the format of the IDs we are using and how should they be generated?

We are putting various "IDs" into JSON documents now being used in the prototype. Use of these IDs (SBI IDs, ScanIDs, PB IDs...) is also identified in the various architecture documents. However, as far as I know there is no agreement on what these are and where they come from. This ticket is to explore this a little before commiting to a solution.

Assume each ID should be unique, assume consistency of approach.

What should be the format? Are they structured? How do we generate them to ensure uniqueness? Are there any architectural implications?