Home Roadmap Mobile Optimization — v2
This is v2, re-measured later the same day, after One Stop Shop shipped. v1 is still here — the selector above this line opens it — and it is not corrected in place, because a proposal quietly edited after a decision was made against it is worth nothing as a record. Two of its numbers were wrong and this page says which.
The home page as a phone meets it: as it stands today, and under four options. Section 01 is the real thing at real size — the controls in those frames are the site's own, painted by the site's own rules. Section 02 draws the same as blocks, because a whole page is nearly four screens long and a frame can only show one. Section 03 answers the question this version exists for: are we closer than we were before One Stop Shop?
Each frame is exactly 390×844: one phone screen. Whatever is inside it is what somebody sees without scrolling, and the bottom edge of the frame is the fold. Scroll inside a frame for the rest of the page.
Scroll sideways for the other three. The controls below are real: .sign,
.cbx, .findem, .stop, .act and
.pill, lifted out of index.html's stylesheet and re-scoped so
they cannot collide with this bench's own classes. Nothing here is wired up
— pressing things does nothing, and the map's dots are placed by hand rather than
by the real projection.
All five at about a twelfth of size, so whole pages fit. Every block is its real measured height; every dashed line is a fold. The outlined block is the attribution — the data-freshness line and the routing credits. Watch it move below the map from Option 0 onward: that is the entire change, it is worth 179px on the number this project exists to lower, and it was invisible until the page was measured twice. Then switch Show to 20 and watch the map, which is the argument for Option 3 over everything else.
Yes, by 57px and a quarter of a screen — and the map went the other way.
Both columns below are the same script, the same default search, run in the same
session: one against the page as it stands, one against
git show 7407317:index.html, the commit that added v1 of this proposal,
served from a second port. That is the only way this question has an answer.
v1 compared its own run to the audit's, found a 25–200px spread and said plainly
it could not fully account for it, because the two runs were not the same search.
Where the three things a reader is looking for actually land. Lower is better in every row.
Costs are estimates and are marked as such. Everything with a number in pixels or screens was measured. The options stack — each one includes the one above it, so Option 3 is Options 0, 1 and 2 with a view switch on top.
Controls under 44×44 at 390px. These are not all the same kind of miss, and
the last row is not a miss at all. This sweep is not the audit's — it
deduplicates by class and drops anything inside an <svg>, so it
counts 12 kinds where _audits/BUDGETS.json records 20. The ratchet in
that file is owned by check-browser.mjs and moves when that script says
so, not when this page does.