We link Stedi's X12 reference. We will not clone it.

Documentation policies are usually invisible until they fail. Ours for EDI is load-bearing enough that we publish it as a feature: original prose about semantics and joins; deep links to Stedi's public X12 reference for segment detail; full dictionary depth only where the underlying content is openly published. This post states the policy, the licensing facts that force it, and the bright line we hold — so a standards-and-compliance reader can audit our surface against it.

The licensing terrain, plainly

X12 is a copyrighted standard sold under license. The organization's IP-use policy is explicit about redistribution: "Direct distribution of X12's copyrighted products by any organization or individual to another organization or individual is strictly prohibited," and "Posting of X12's copyrighted products on another organization's public website is strictly prohibited" (x12.org/products/ip-use — quoted here within the policy's own allowance for short citation). What the policy permits without a license is internal use with citation and short citations for comment, teaching, and research. Even companion guides may be distributed to trading partners only after explicit permission.

The published license tiers (x12.org/products/licensing-program):

Read that last row carefully, because it is the whole story: no fixed-price tier grants republication. The right to serve X12 dictionary content on a public website is not something you buy off a rate card; it lives in negotiated partner agreements. Any "free X12 reference" you find on the open web is either Stedi's — published under their own arrangement — or it is someone's future takedown notice.

The bright line: facts and methods vs expression

Copyright does not protect facts or methods; it protects expression. Our documentation lives entirely on the safe side of that line:

• Structure in our own words. What a despatch advice is for, how its packing hierarchy nests, which fact maps to which EPCIS field — original prose, ours. The 856 mapping post is the working example: the hierarchy explained without reproducing a single dictionary entry.
• Self-authored fixtures. Every wire example in our corpus is synthetic. No partner implementation guide, no example library content, reproduced.
• Self-authored grammars. The parsing engine's structural data files are our own work, built from corpus observation and government-published implementation conventions, each with a provenance ledger. Element descriptions are in our words or omitted — the engine needs structure, not prose.
• Segment-level detail is a link. Where a reader needs element-by-element depth, we deep-link Stedi's X12 reference with visible credit. It is live, free, and stable — and linking the canonical reference machines already know is honest answer-engine practice as well as honest IP practice. Credit where the work is: Stedi built the reference the whole developer-EDI world reads. We are not going to launder it.

The asymmetry we exploit instead of resenting

The international document set has the opposite posture. UNECE publishes the UN/EDIFACT directories openly (unece.org/trade/uncefact/unedifact/download). So the EDIFACT side of our house goes dictionary-deep under the same pin discipline as the GS1 artifacts on the conformance page: the business-transaction layer's committed provenance ledger (vendor/untdid/PINS.json) governs the directories per file — sha256, source URL on unece.org, retrieval date, pins changed only by journaled human ruling — and every directory fact renders from that committed artifact. The result is a document side treated at a depth the X12 side legally cannot get from us: the full DESADV→EPCIS mapping, CPS packing tree and all.

One consequence worth naming for RFP season: when you evaluate vendors in this space, ask each one where their X12 documentation came from. A vendor republishing dictionary content they cannot license is showing you their diligence posture on every other standard they touch — including the GS1 artifacts your conformance story depends on. Ours is checkable: pinned sources, provenance ledgers, and a public policy you are reading now.

The policy, as a checklist

1. X12 semantics and joins: original prose, always.
2. X12 segment/element detail: link Stedi, credit visibly, copy never.
3. Wire examples: synthetic, self-authored, license-tagged in the corpus.
4. EDIFACT and GS1 artifacts: governed by committed sha256 pin ledgers, changed only by journaled ruling.
5. No exception for convenience, marketing, or "everyone does it."

That is the whole policy. It costs us some documentation surface on the X12 side, and it buys something worth more: every page we serve is a page nobody can make us take down. The same discipline is why the differential harness manufactures its ground truth from independent parsers and licensed-clean fixtures rather than from spec text we could not ship — the IP posture and the testing posture are one decision, made once.

Browse the open side now: the DESADV→EPCIS mapping is live, at full depth where the license allows. The thesis it feeds — documents as event vocabulary — is here.