From 9ff0183844b8a35375dd5a3dfa0c4b08c90c80ed Mon Sep 17 00:00:00 2001 From: Yorke Rhodes III Date: Tue, 25 Aug 2026 16:52:18 -0400 Subject: [PATCH] Close UPD-017: the symbol fallback, not 139 conversions Records what was actually done and why the original proposal was wrong, including the part of the acceptance criterion that is deliberately not met. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --- UPDATES-NEEDED.md | 68 ++++++++++++++++++++++++----------------------- 1 file changed, 35 insertions(+), 33 deletions(-) diff --git a/UPDATES-NEEDED.md b/UPDATES-NEEDED.md index 970c3a358..d007c1894 100644 --- a/UPDATES-NEEDED.md +++ b/UPDATES-NEEDED.md @@ -415,7 +415,7 @@ family ships. ### UPD-017 - Decide how UI symbols are drawn -**Priority:** Medium +**Priority:** Medium — **done.** The nav's dropdown caret was the character `▾` (U+25BE). `next/font` subsets Space Mono to latin, latin-ext and vietnamese and emits a matching @@ -423,38 +423,40 @@ Space Mono to latin, latin-ext and vietnamese and emits a matching permitted to use Space Mono for it and substituted whatever the visitor's OS provides. The result sat next to Space Mono text in a different typeface, at a different optical size, on its own baseline. Reported as a font inconsistency in -the header on 2026-08-25 and fixed there by drawing the caret as an inline SVG. - -**The same is true of every other symbol used as an interface affordance.** The -subsets cover `U+2191` and `U+2193` but not, among others: - -| Glyph | Code point | Used in | -|---|---|---| -| `→` | U+2192 | `Contact →`, `View profile →`, cohort and archive links | -| `←` | U+2190 | back links, poster rail | -| `↗` | U+2197 | external links | -| `▶` | U+25B6 | launch-demo buttons | -| `▸` | U+25B8 | cohort and chart markers | -| `✕` | U+2715 | close buttons | -| `●` `◦` `─` `＋` `📖` | various | explorers, diagrams, book trigger | - -None of these is broken today — they render from a fallback, and because they -*all* fall back they are at least consistent with each other. But the typeface -is whatever the visitor's platform supplies, so the site looks different on -Windows, macOS, Android and Linux in a way no one here controls, and a missing -glyph would show a tofu box. - -**Proposed update:** Decide one approach and apply it in a single pass rather -than piecemeal — converting one arrow at a time is what would create a real -inconsistency, since the converted one would no longer match its neighbours. -Either draw them as inline SVGs (as `LinkedInLink`, `ThemeToggle` and now the -nav caret already do), or add an explicit symbol fallback stack so the -substitution is at least a deliberate choice. Worth a look at -`docs/DESIGN-SYSTEM.md` at the same time, which documents typography but says -nothing about symbols. - -**Acceptance:** Interface symbols render identically across platforms, and the -rule for adding a new one is written down. +the header on 2026-08-25. + +The same was true of every other symbol used as an interface affordance: `→` in +`Contact →` and `View profile →`, `←` in back links, `↗` on external links, `▶` +on launch buttons, `▸`, `✕`, `●`, `◦` and `＋` — about 139 rendered occurrences +across 43 files. + +**Done, in two parts.** + +The caret is drawn as an inline SVG, because it also needed to be a specific +shape at a specific size: the substituted glyph and then a too-small filled +wedge both read as a full stop at 14px. It is now a chevron. + +Everything else was fixed by **correcting the fallback chain rather than +converting 139 call sites**. The body face is the monospace, but the chain read +`Arial, Helvetica, sans-serif`, so every glyph outside the subset was drawn by +Arial — a proportional face inside monospace text. Chrome's +`CSS.getPlatformFontsForNode` on a back link reported `Arial (1 glyph)` beside +`Space Mono (31 glyphs)`; after naming a monospace fallback in `globals.css` and +in the `next/font` config, the same probe reports `Cascadia Mono`. That also +removed a latent defect: had the webfont ever failed to load, the entire site +would have rendered in a proportional face. + +Converting the glyphs to SVG was the original proposal and would have been the +wrong call — an enormous diff, and impossible for the glyphs that live inside +content strings in `src/content/`. + +**Acceptance:** met for consistency — symbols now render in a monospace face +alongside monospace text, and degrade to a monospace stack rather than to Arial. +Not met for pixel-identical cross-platform rendering: `ui-monospace` leads the +stack, so each platform still uses its own UI monospace. That is a deliberate +trade, and the remaining lever would be self-hosting a font that covers these +codepoints. `docs/DESIGN-SYSTEM.md` documents typography but still says nothing +about symbols, which is worth adding if anyone codifies this further. ## B. Draft requests to document the maintenance workflow - review before sending