Smartsensor Unique ID Allocation Scheme

[nikitakit]

Smart sensors should be allocated unique 64-bit (8 byte) IDs. I will be updating this with proposed schemes for allocation

Potential design considerations:

• Let's allocate some of the ID space for non-PiE vendors, so other people can choose to adopt the protocol
• Type/subtype allocation in order to allow certain types of sensors to enumerate faster
• Different computers will be used to program sensors. The IDs should not conflict between the different computers
• We do production in a massive push, once a year. We should be able to resume each year's numbering from where the last year left off. If we forget where the previous year's numbering left off, we will instead have to throw away huge chunks of address space because we suspect that it may have been used.

[nikitakit]

My proposal:

• 4 bytes: Vendor ID.
• 2 bytes: Vendor-specific sub-id
• 2 byte: index (increment numerically from 0)

Vendor IDs are:

• 0: Debug only (for assigning to devices during prototyping. Do not use publicly)
• 1: PiE
• All others: reserved for future use

Sub-id:

• 0: Common use laptop 0
• 1: Common use laptop 1
• ...
• ?: Common use laptop 0, again, once it uses all ids available in the last 2 bytes
• ...
• 0xFF00: Casey's computer
• 0xFF01: Robert's computer
• ...

(since we may end up using our own computers during production, let's work this into the plan)

This makes the tradeoff of not doing the category/subcategory for enumeration, but instead works on the problem of ensuring that all IDs are unique, across different computers and years of PiE.

[nikitakit]

Suggestion: change to

• 1 byte vendor ID. (0=debug, 1=PiE)
• 3 bytes reserved for future use. Must be set to all 0
• 2 bytes computer id
• 2 bytes index

[ArcaneNibble]

final suggestion accepted, sounds good