How 4D-ID is built, and what we considered

A candidate standard earns trust when you can see not just what was chosen but what was rejected and why. These are the current design proposals behind 4D-ID, each a choice to test against real alternatives.

The current proposal covers persistent identity, hierarchy, pose and time records, aliases, representations, provenance, and lifecycle. A rule belongs in the candidate standard only when independent implementations need it to interoperate.

The decisions

1. An identity layer, not a coordinate system. We considered defining a new unified frame that everything converts into. Every domain already has a native frame it will not abandon; a new one is just a sixth silo. What none of them has is a stable name for a thing. So we add a name, not a map.
2. A discrete global grid at the root. A grid cell can provide an anchor, partition key, and query/index key. It is not the identity by itself, and localization, route planning, and navigation remain external responsibilities.
3. A strict tree, not a graph. We considered letting things have several parents. Trees keep every position unique and every traversal cheap, and they cover every concrete use we could find. Multi-parent graphs need disagreement-resolution rules nobody has specified well, so we deferred them.
4. The name records where it was born, not where it is now. Identity must be stable; location must be current; those are different needs. Putting the birth-place in the name and the current place in the state is what lets any name resolve in two hops with no global directory.
5. One fixed rule for which way is up. Graphics systems are Y-up; geospatial systems are East-North-Up. We considered letting each implementation declare its own mapping. Ambiguity there is the single most common cause of mirrored, drifting, upside-down shared content, so we fixed one mapping and allow no other.
6. Availability over consistency when the network breaks. We considered requiring global agreement before anything can be written. A robot in a tunnel and a headset in a basement have to keep working. So each side keeps operating, staleness is marked, and the sides reconcile when they reconnect.
7. Signed proof, not a ledger. Some uses need cryptographic proof that a thing was at a place at a time. We considered a blockchain. A signature from a known key provides the proof without the consensus cost or the token economics. Ledgers are allowed as a deployment choice, never required.
8. Privacy and security as requirements, not considerations. A stream of where everything is, with history and prediction, is a surveillance system unless the standard says otherwise. So pseudonyms, scoped disclosure, retention limits, consent, and audit are normative, and privacy claims are demonstrated by test, not asserted.
9. Automation proposes; authority promotes. An AI or a sensor can suggest that a thing exists, or propose a change. It cannot make that authoritative on its own. Inferred content enters as evidence and becomes real only through an authorized promotion. This keeps an AI from silently turning a guess into ground truth.
10. A candidate identifier is free to create. Anyone can generate a syntactically valid 4D-ID without a central registry. A UUID also provides uniqueness; 4D-ID adds standardized resolution, aliases, hierarchy, representations, state and time, provenance, merge/split lifecycle, and historical continuity. Neither token alone establishes what real-world entity it refers to.

What we have not decided

A candidate standard is honest about its open questions. Deferred, on purpose, until real use makes the right answer clear: the exact wire encoding; a reference scheme for pseudonymity; a policy language for scoped disclosure; which grid variant to mandate; the algorithms for deciding two observations are the same thing; and multi-parent graphs. These are written down as open, not hidden as finished. Contribution is welcome on all of them.

The full reasoning, with every alternative, lives in the specification's design-rationale annex. This is the short version, and it is the version that answers the question every serious reader asks first: why this, and not that.