Attributions

Text and specifications drawn from Wikipedia are used underCC BY-SA 4.0. Entity identifiers and image resolution come from Wikidata (CC0). Photography comes from Wikimedia Commons, licensed per file and credited on the page it appears on.

Photography is credited on the photograph itself, not listed here. Every image carries a smallibutton giving its author, licence and source page. That is deliberate: a single list of every image on the site would run to many thousands of rows and separate each credit from the picture it belongs to, which serves neither the reader nor the photographer. This page stays for the source works the site's data is drawn from.

Manufacturer names, model names and logos are the trademarks of their respective owners. They appear here only to identify those manufacturers' vehicles, unmodified, and nothing on this site implies any endorsement, partnership or affiliation.

Markey's wind tunnel is our own implementation, but its structure follows Aeolus T-1, a browser lattice-Boltzmann solver whose author gave permission to use or modify it. The arrangement of the D3Q19 velocity set with opposite pairs adjacent, the deviation-form storage that keeps a float32 texture viable, the packing of nineteen distributions into five render targets, and the discipline of generating the GPU shaders from the same table the CPU reference uses: all of that is theirs.

So is a warning that saved a day of work: holding the TRT magic parameter fixed while lowering viscosity to raise the Reynolds number drives the antisymmetric relaxation rate toward zero, and the solver quietly diverges. It is documented in ourlattice.ts with credit, because the next person to touch it will need it.

The underlying methods (Qian–d'Humières–Lallemand D3Q19, Ginzburg–Adler two-relaxation-time collision, Smagorinsky subgrid viscosity, momentum-exchange force integration, and the Clift–Gauvin drag correlation used to validate against) are published science and belong to no one.

A single site-wide bibliography was the first thing built here, and it was the wrong shape. With four cars it held eight entries; at a thousand cars it would hold thousands, and a reader checking one figure on one car page would have to search the whole list to find the two documents that page actually depends on.

So citations work the way they do in print. A claim in the prose carries a number, [1], and so does every sourced figure in the specification tables. Both point at a References section at the foot of that page, listing only what that page cites, with the revision and the date each source was read at. It is the same reasoning that puts a photograph's credit on the photograph.

The bibliography those numbers resolve into is a single checked file,src/data/references.json. A citation pointing at a key that is not in it fails the build rather than rendering a number with nothing behind it.