Repository

github.com/ilang-ai/iml-protocol · concept DOI 10.5281/zenodo.22823285 · MIT

What it is

IML (I-Lang Machine Layer) renders the declaration layer of I-Lang v4.x as fixed-width codes: 49 codes covering the declaration set, each derived from the canon rather than invented beside it, with a digest so a receiver can check which table it is reading. It is meant for the leg between two machines, where a human reader is not the audience and a stable width is worth more than readability.

It is experimental and not part of the canon. The canon stayed untouched when IML was built; what moved was the declaration layer’s representation, not its definition. Version 0.5.1 is the current and, for now, the final release: the line reopens when there are real users or when the canon settles the field syntax it would depend on.

What it does not claim

IML is not lossless compression and it does not save tokens — in the current tables an IML document is longer than the I-Lang document it encodes. That is stated in the repository’s own README, and it is stated here for the same reason: the machine layer exists for determinism at the boundary between agents, not for a smaller bill.

The canon itself: ilang-ai/ilang-spec, concept DOI 10.5281/zenodo.21821452. Conformance measurement of the canon: I-Lang Conformance Results v1.