Skip to content

sorted-json-v1.md

Latest commit

 

History

History
52 lines (32 loc) · 1.82 KB

sorted-json-v1.md

File metadata and controls

52 lines (32 loc) · 1.82 KB
Raw

sorted-json-v1

Status: SHIPPED prototype canonicalization profile

sorted-json-v1 is the current canonical JSON byte rule used by the browser evidence and witness-bundle prototype.

Rule

  1. Recursively sort object keys by Unicode code-unit order as produced by JavaScript Object.keys(value).sort().
  2. Preserve array order exactly.
  3. Omit object properties whose JavaScript value is undefined before serialization.
  4. Serialize the resulting JSON-compatible value with compact JSON equivalent to JavaScript JSON.stringify.
  5. Encode the resulting string as UTF-8 before hashing or signing.

Cross-language safe domain

The published golden vectors currently cover:

  • objects;
  • arrays;
  • strings including Unicode;
  • booleans;
  • null;
  • integers within JavaScript's safe integer range.

The current profile does not claim fully specified cross-language behavior for every IEEE-754 edge case, including non-integer numeric formatting, -0, NaN, Infinity, or values outside JSON's data model.

Implementations should reject or separately version unsupported values rather than silently inventing incompatible canonical bytes.

Golden vectors

See:

protocol/browser-canonicalization-vectors.json

Each vector publishes:

  • input JSON;
  • exact canonical UTF-8 JSON string;
  • SHA-256 of those bytes.

Versioning rule

A future canonicalization rule that changes canonical bytes must use a new canonicalization identifier. It must not silently redefine sorted-json-v1.

Why this is explicit

Two runtimes can both say "sorted JSON" and still disagree on bytes. Cryptographic interoperability requires the bytes—not merely the data structure—to match.

Claim boundary

This profile is a project-defined prototype contract. It is not an IETF, W3C, WHATWG, ISO, or other externally ratified canonicalization standard.