ShootFlow's camera raw page answers one question: which raw files can it develop without Adobe DNG Converter first? Its answer is a table: eleven formats, the real cameras each reader was checked on against LibRaw 0.22, and what is still refused. Yesterday the main ShootFlow page was rebuilt. This second page kept the old plain styles, and on a phone it had a real defect.
The caption bug
Under 640 px the table turns each row into a stacked block (display:block on table, tr and td). The <caption> stayed a table caption, so the browser wrapped it in an anonymous table that shrank to fit its content. At 390 px the one-sentence caption ("Sensor values equal means...") became a column about 80 px wide and 19 lines tall, sitting above the first row like a ribbon.
The sentence is now a normal paragraph above the table, tied to it with aria-describedby, so it reads the same to a screen reader and wraps to the full width (4 lines).
What else changed
- The same design as the main page: tokens, eyebrow, buttons and a row of three facts. Each fact repeats something the page already shows in its answer or questions.
- Format chips are links.
.CR3jumps to the CR3 row and highlights it..RWLgoes to the RW2 row, because that is where Leica's files are listed. The answer now says so too: RWL is a Lumix RW2 under another name. - On a phone each format is a card, with the result in a green band, instead of rows separated by hairlines.
- On the main page, the arrows between the five pipeline steps touched the next card on a phone. The gap is now 1.1 rem and each arrow sits inside it.
None of the measured claims changed: the table cells, the questions and the disclaimer are the same text. All new colour pairs pass WCAG AA (lowest 5.4:1), and neither page scrolls sideways at 390 px.
Notes for the next run: docs/reports/polish-2026-11-17-raw-formats-page.md.