0.8.0 — the photo was smaller after you clicked it

08:11. The report went roughly: in the stream the photo is wider than after you click through to the detail, that makes no sense. And indeed it made none.

The numbers: the entry list rendered in a column 768 pixels wide, a single entry page in 672. So the reader clicked a thumbnail in order to see the photo and received it a hundred pixels narrower. No test catches that. Tests check whether the image is there. Nobody writes an assertion saying “the image is larger than it was on the previous screen”, because it sounds absurd — until you see what it looks like when it is not.

The fix consisted of separating two things that had been sharing one container. The text column stays narrow, because 672 pixels is about sixty-five characters to the line and going past that damages reading. The photo leaves that column and goes to the full width of the page.

And here is a detail I strongly advise against skipping. Before setting the width to a hundred per cent, I measured the actual dimensions of the photographs in the existing site: median 1536 pixels, half of them below 1600. On a 2560-pixel monitor, a literal hundred per cent would mean stretching most of them beyond the resolution of the source. The image would be bigger and noticeably softer — so the change would have worsened precisely the thing it was meant to improve.

Hence the ceiling of 1600 pixels. On a typical laptop the photo still fills the page width, and on a very wide screen it stops exactly where the material runs out.

“Make it full width” was the right instinct. “A hundred per cent” was not the right execution of that instinct. The difference between the two is fifteen minutes with a console and an image library.