> For the complete documentation index, see [llms.txt](https://hetcreep.gitbook.io/hetcreep-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://hetcreep.gitbook.io/hetcreep-docs/contributing.md). # Contributing Read this section before you write anything. This is not a codebase, and the usual contributing file would waste your afternoon. This document is licensed **CC BY-NC-ND 4.0** and has a single owner. **NoDerivatives** means nobody else may publish an adapted version of it, and a pull request that edits the text is an adaptation. So the ordinary open-source loop — fork, edit, PR — is not the model here, and there is no point pretending otherwise. Sending one is not rude; it is just something that cannot be merged. What **is** wanted is narrower and considerably more valuable: **evidence that something in it is wrong.** The standard makes falsifiable claims about vendor behaviour, store requirements, and API guarantees. Every one of those is a hostage to a vendor who can change it without telling anyone. A document that was correct the day it was published goes quietly wrong while every word of it stays exactly where it was. Catching one is the most useful thing a reader outside this repository can do, and it is the thing this project cannot do for itself at scale. *** ## Three things worth sending ### 1 · A correction, with evidence The unit of a useful report is not "this is wrong". It is: * **the location** — the rule ID or the register row, and what the document says today, **quoted**; * **the source** — a URL to the vendor's or standards body's own page. Not a tutorial, not a forum answer, not a second-hand document that cites the first one; * **its class** — specification · vendor documentation · tool default · recommendation. These do not weigh the same, and a report that does not say which one it is has skipped the hard part; * **the exact wording you found there**, quoted rather than summarised; * **how the two differ**, in a sentence; * **the date you fetched it**, because the next reader needs to know how old your check is. **A report with no source cannot be acted on.** Not because it is unwelcome — because there is nothing to act on. "The Unity limit is different now" leaves the owner to do the whole job from scratch and to guess which page you were looking at. "Unity Manual, *Import a texture*, fetched 2026-08-11, says X, the register says Y" is a decision someone can make in two minutes. A verified defect is worth more than a polite one. If a number in here is wrong, say so bluntly and show the page. ### 2 · A named counter-exemplar For a `B-ext` value — an external convention with real precedent — the standard's own rule is that it moves **only with a named counter-exemplar**. Not "most projects do it differently". Not "in my experience". A real, shipped, nameable thing that does it another way, named specifically enough that someone else can go and look at it. That rule exists because "everyone knows" is how the conventions this document is trying to disentangle got established in the first place. ### 3 · A re-measurement `LICENSE`'s scope section says it directly: measurements, numbers, and facts are not copyrightable, and **anyone may independently re-measure the same public corpora and publish their own numbers.** That is expected. It needs no permission and you should not ask for any. The method is published in the standard precisely so it can be re-run rather than trusted. Run it, publish your figures in your own repository under your own name, and send a link. Independent re-measurement is not a threat to a standard; it is the only way one earns the right to be believed. If your numbers disagree with the published ones, that is a correction under item 1 and the more interesting outcome of the two. *** ## What will not be accepted * **Rewordings, restructuring, and style.** The expression is the part the licence actually protects, and it is the owner's. * **New rules proposed without evidence.** A rule with no source is a Layer C value with ambitions, and promoting one to spec to make a table look complete is the specific failure this document was written against. * **Anything that requires your text to be incorporated.** This is an **open question, not a policy**: `LICENSE` says what the world may do with this document and says nothing at all about what rights the owner would hold over text a contributor sends in. That gap is not something this file gets to fill by inventing a CLA. Until a lawyer settles it — the same lawyer `LICENSE` already says has not looked at any of this — **send evidence and let the owner write the words.** It sidesteps the question entirely and costs you nothing, because the valuable part of your report was never the prose. *** ## Where a report goes **Open an issue.** That is the route to use for a correction, a counter-exemplar, or a re-measurement. It is public, it is harder to lose than a message, and a report anyone can read is a report anyone can check. For a **licence request** — commercial use, adaptation, redistribution of the document — either an issue or a direct approach through **github.com/HetCreep** works. `LICENSE` names both. > An earlier revision of this section said there was no issue tracker, because there was no public remote, and instructed that it "gets updated in the same commit that adds the remote". The remote was added and this section was not updated in that commit. It is recorded rather than quietly replaced, because a file that predicts its own staleness and goes stale anyway is a better argument for checking than any rule about it. ## What happens to it A correction with evidence is written into [`ERRATA.md`](/hetcreep-docs/errata.md) as **reported**, whether or not anyone agrees with it yet. The owner then rules: **verified**, **rejected**, or **held for the next version**. **Entries are never removed, including rejected ones.** A rejected report keeps its evidence and its reasoning, which is what stops the next person re-reporting the same thing. If yours is rejected you will be able to read exactly why, and if the reasoning is wrong, that is itself reportable. Two things worth knowing before you expect a fast fix: * **A released version does not change.** A defect found in `1.0.0` is corrected in the next version, not edited out of `1.0.0`. Someone conforming to `1.0.0` has to be able to open it and find the same document. See [`GOVERNANCE.md`](/hetcreep-docs/governance.md). * **Being listed in `ERRATA.md` is not being fixed.** It is being on the record, which is a different and slower thing, and the register says which state each entry is in so nobody has to guess. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://hetcreep.gitbook.io/hetcreep-docs/contributing.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.