Licence#

Touchstone is Apache-2.0. The full text is in LICENSE, and the copyright notice is in NOTICE.

Two clauses did the choosing. Apache §3 grants patent rights expressly, with defensive termination; MIT has no patent grant at all. That matters here more than it does for most projects, because the author of this standard is a commercial competitor to the labs most likely to adopt it, and a lab's counsel is entitled to ask in writing whether the author can sue them later. Apache §6 then withholds trademark rights, which is why conformance below is something you state rather than something you are granted.

The rest of this page is generated from CONTRIBUTING.md in the repository, so the terms on the website and the terms in the tree cannot come apart.

Licensing#

• Apache-2.0 - the code (scoring.mjs and any other source), the JSON Schemas in schemas/, the conformance vectors in test/, the rubric data, and the normative specification text in RUBRIC.md.
• CC BY 4.0 - explainer prose: documentation that describes the standard rather than defining it.

Where a file doesn't say which one applies, it is Apache-2.0. The full text is in LICENSE.

Trademarks are not licensed. Apache-2.0 §6 grants no rights in the licensor's trade names, trademarks, or service marks, and nothing here grants them separately.

Against that, one commitment. You may state that your implementation conforms to Touchstone, naming the version it passes: conforms to Touchstone 0.1. We will not assert our marks against an accurate claim of that form. Name the version. Conformance is defined per version — it means passing that version's vectors in test/ — and a bare "Touchstone conformant" with nothing behind it is a claim we would treat as inaccurate.

This holds even if we later run a paid certification programme. Certification would attest that we checked. It would never become the price of accurately reporting a result you can reproduce yourself by running the vectors.

You do need written permission to use the Touchstone name as the name or branding of your own product, or in any way that suggests we endorse, certify, or stand behind it. Ask first. Open an issue.

The licensing commitment#

Four things are covered: the specification (RUBRIC.md), the rubric data, the JSON Schemas, and the conformance vectors. They remain permanently under a licence that never imposes obligations on your work merely because that work conforms to, or tests against, this standard. In practice: Apache-2.0, MIT, BSD, or CC BY 4.0. That rules out non-commercial and no-derivatives terms, and it rules out copyleft and share-alike terms just as firmly. The schemas are meant to be embedded in your product and the vectors are meant to run inside your test suite. A licence that attached conditions to your code or your internal documentation for doing either would defeat the point of publishing them at all.

This commitment may be broadened but never narrowed. It is the one part of this document the amendment rule under Governance does not reach, and it cannot be changed by the proposal process below.

The reference implementation (scoring.mjs) is deliberately not covered. Its licence may change, with twelve months' notice and a final release under the prior terms. Notice means a dated entry in this file, repeated on the releases page and the package registry listing. Contributions are accepted on the ordinary terms throughout the notice period, and the final release under the prior licence is of the tree as it stands at the end of that period, not as it stood on the day notice was given.

A promise is worth what the party behind it is worth. This commitment is made by Grailology, LLC. It does not bind an acquirer and it does not survive the company's wind-down.

The Apache-2.0 grants already made are irrevocable. Every version published to date can be used, modified, redistributed, and forked by anyone, forever, regardless of anything we later decide and regardless of who ends up owning Grailology. That part does not depend on trusting us. If this project goes somewhere you won't follow, that is the floor you land on. It is the one thing here that needs no good faith from us at all.

None of this obliges us to keep working on Touchstone. We could hold every published version open exactly as promised, stop development, and let the standard go stale, and every word above would still be true. That is the honest risk of adopting a standard with one maintainer. You should price it in rather than take comfort from this section. The half we can close, we close. If we publish a successor standard, it goes out under these same terms. We will not freeze this one and reintroduce it as a product.

That is the entire commitment. It is deliberately narrow. We would rather give you a promise we can be held to than a bigger one we can't.