Skip to content

How Canon verifies

Canon answers two separate questions from raw Radiant transaction evidence, and never merges them.

1. Is the relationship valid?

A Glyph declares collection (in) and creator (by) relationships inside its metadata. Metadata is just bytes — anyone can write anything there. What cannot be copied is the on-chain authorization:

  • Direct:the related collection or creator token itself participates in the Glyph's reveal transaction — its unique reference is re-created in an output, which the chain only permits when the token was actually spent as an input.
  • Delegated:the creator minted a delegate token whose creation required control of the collection/creator references. The Glyph's commit transaction consumed that delegate and its reveal provably burned it. Canon retrieves the delegate's creation transaction and checks the whole path.

Canon parses the actual commit, reveal and delegate transactions — script by script, opcode by opcode — and additionally checks that the commit output committed to the exact metadata bytes the reveal published. Indexers are used only to find transactions, never to judge them.

2. Is it the reference you expected?

The chain can prove a relationship to reference X. It cannot know whether Xis the collection you meant. The expected reference must come from a trust anchor outside the chain: the creator's website, a verification link or QR code they published, or a receipt received through a trusted channel. Canon compares the proven relationships against your expectation and reports a match, a different reference, or an unproven claim — nothing more.

What the colors mean

  • Green — a cryptographic or transaction-evidence check completed successfully.
  • Amber — externally supplied expectations, unconfirmed evidence, missing chain data, or mutable metadata.
  • Red — a definite failure: an expected-reference mismatch, an unauthorized declared relationship, malformed data, or a media digest mismatch.

What Canon never says

Canon does not label anything “authentic”, “official”, “trusted” or “counterfeit”. A valid relationship to a different reference may be entirely legitimate — it is simply not the one you asked about. Missing network data is reported as unavailable evidence, never as invalidity.