This page describes v1.13. Verify whether it is the current manual at /version.json
How to reason about it
- The format pins its image, inspection, profile, and codebook tags; 1304-pixel width; palette; page geometry; fixed-Huffman PNG encoding; stored ZIP grammar; page order; and manifest bytes.
- Inputs are bounded to 1 MiB of source, 16,384 reader nodes, reader depth 256, and 32 pages; malformed length and offset fields are rejected before unsafe allocation.
- Inspection exposes format/profile identifiers, page count, exact source length and identity, plus projection, Core, payload, page-set, and per-page commitments.
- Reed–Solomon parity detects changed payload symbols but performs no repair. A valid payload must also regenerate to exactly the supplied PNG or ZIP bytes.
- The round-trip laws are
decode(encode(source)) = sourceandencode(decode(image)) = imagefor every accepted image.
A common mistake
Canonicality proves only that the bytes are one exact Lispex Image representation. Source identity, signatures, trust policy, request binding, Native re-execution, and a local gate remain separate boundaries.
Current boundaries
- Do not infer watermarking, steganography resistance, visual robustness, error correction, confidentiality, authorship, or malicious-code safety from a valid image.
Keep going
Use this page while writing code. For a guided explanation, return to the syntax map or the corresponding manual.