Editorial standards

What every article has to meet

Each piece starts from primary material: the specification, the kernel source, the vendor documentation, or a measurement taken on our own hardware. Before it is published it is checked against a fixed set of standards, and a draft that misses any of them is held back rather than trimmed to fit.

• A minimum length, with a clear opening, section headings, a conclusion, and a references list that points at the sources actually used.
• No more than 40% title overlap or 80% body overlap with anything already published here, which is how we keep the same subject from being covered twice under different words.
• At least two contextual links to related material on this site, so a reader can follow a mechanism across articles.
• English throughout, and a lead image that belongs to the article.

On a scheduled publishing day with nothing that meets these standards, nothing is published. We would rather skip a day than fill it.

What we measure ourselves

An explainer restates knowledge that already exists. To be worth your time, this site also has to produce something that does not exist anywhere else, so we run benchmarks on the server that hosts this site and publish the numbers along with the complete source code, the exact method, and the machine specification.

You can disagree with a method, and you can run any of them yourself: see the benchmarks.

Corrections

Technical writing gets things wrong. If you find an error — a misstated mechanism, a wrong number, a claim that does not survive contact with the source — tell us through the contact page and we will correct the article and say what changed.

We publish no sponsored articles and no paid product placements. Where advertising appears, it is clearly marked as advertising and has no influence on what we write.