/* OPULENT AURA — SELF-HOSTED FONT LAYER v0.2 Produced by Nasaq, Design-System Engineer · 2026-07-31 BINDING RULES [Typeface Ruling, Haneen, §Implementation]: · self-host WOFF2 ✅ SATISFIED as of this version · NO third-party font CDN at serve time ✅ nothing here leaves the palace · version-pin the adopted release ✅ recorded in tokens-v0_2.css · ship the OFL licence beside each family ✅ fonts//OFL.txt, four of four · never modify and redistribute under the reserved font name · acquisition from Google Fonts / Fontsource official releases only «The no-CDN rule is not only consistency. It is the palace rule: nothing on a house surface depends on a call leaving the palace.» — Haneen ── FORMAT: WOFF2 FIRST, TTF FALLBACK ── v0.1 of this file shipped TTF-only, because no WOFF2 encoder existed on this palace. Qudra built one — offline, nothing uploaded — and converted the families. The browser takes the first format it understands, so WOFF2 is listed first. ── WHAT THE 13 DECLARED FACES ACTUALLY WEIGH (measured, not estimated) ── TTF 3,855,436 B (3.68 MB) WOFF2 1,000,320 B (0.95 MB) saving 74.1% This is the figure that matters for the W6 speed budget — the weight a surface actually ships, not the size of the kit on disk. ── HOW THESE FILES WERE VERIFIED, BY ME, BEFORE BEING BUILT ON ── 1. wOF2 signature on all 24 files of the four ratified families — 24/24 PASS. 2. Round-trip decompression of all 24 with the same engine — 24/24 succeeded, and each output matched the size its own header declares, exactly. 3. Metrics of every round-tripped face compared against its source TTF — unitsPerEm · ascender · descender · numGlyphs — 24/24 IDENTICAL. So the files are complete and faithful; the conversion lost nothing. ── THE 0.3–0.6% SIZE ANOMALY: NOW SOLVED, AND SOLVED BY MEASUREMENT ── Each reconstructed sfnt is 0.3–0.6% larger than its source TTF. My first hypothesis was the short→long `loca` conversion. I TESTED it and it FAILED — no file changed loca format. I recorded the cause as UNDETERMINED rather than invent a second guess. Qudra then measured it, and I reproduced his measurement independently: CAUSE: WOFF2 reconstruction aligns every glyph to a 4-byte boundary. The source packs its glyphs contiguously. The increase is the sum of that padding. glyph offsets 4-byte aligned, source -> reconstructed: Cormorant Regular 679/2831 (24.0%) -> 2831/2831 (100%) Amiri Regular 918/6624 (13.9%) -> 6624/6624 (100%) Arsenal Regular 492/990 (49.7%) -> 990/990 (100%) Almarai Regular 249/531 (46.9%) -> 531/531 (100%) padding predicted from the source's own glyph lengths vs the delta observed: Cormorant predicted 4,101 observed 4,100 residual -1 Amiri predicted 3,003 observed 3,000 residual -3 Arsenal predicted 812 observed 796 residual -16 Almarai predicted 516 observed 516 residual 0 Residuals are table-level padding, not glyph-level. And it explains the thing that looked wrong: Amiri has 6,624 glyphs and a SMALLER delta than Cormorant's 2,831 — because the right denominator is glyph LENGTHS, not glyph COUNT. Bytes-per-glyph was a misleading ratio. NOT LOSS: metrics and glyph counts identical across all 24 faces. ⚠️ BUT THE ANOMALY HAS A CONSEQUENCE FOR PINNING, AND IT IS NOT COSMETIC: reconstruction does not return the source byte-for-byte, so "this WOFF2 derives from a pinned TTF, therefore it is pinned" DOES NOT HOLD. Every shipped WOFF2 carries its own SHA-256 in tokens-v0_2.css, beside the encoder that produced it. PATHS assume this file sits beside ./fonts/. Adjust the prefix once, here, per surface. */ /* ═══════════════════ ROLE: IDENTITY · LATIN — Cormorant Garamond 4.001 ═══════════════════ */ @font-face { font-family: 'Cormorant Garamond'; font-style: normal; font-weight: 300; /* Light — the VIS display weight */ font-display: swap; /* [W6] with metric-compatible fallback */ src: url('./fonts/Cormorant-Garamond/CormorantGaramond-Light.woff2') format('woff2'), url('./fonts/Cormorant-Garamond/CormorantGaramond-Light.ttf') format('truetype'); } @font-face { font-family: 'Cormorant Garamond'; font-style: normal; font-weight: 400; font-display: swap; src: url('./fonts/Cormorant-Garamond/CormorantGaramond-Regular.woff2') format('woff2'), url('./fonts/Cormorant-Garamond/CormorantGaramond-Regular.ttf') format('truetype'); } @font-face { font-family: 'Cormorant Garamond'; font-style: italic; font-weight: 300; font-display: swap; src: url('./fonts/Cormorant-Garamond/CormorantGaramond-LightItalic.woff2') format('woff2'), url('./fonts/Cormorant-Garamond/CormorantGaramond-LightItalic.ttf') format('truetype'); } /* ═══════════════════ ROLE: SUPPORT · LATIN — Arsenal 2.000 ═══════════════════ Four styles total. [RD1] records the trade honestly: we lose flare INTENSITY, the direct price of the Founder's no-commercial-licence rule. Function is fully replaced — x/em 0.500 carries every T3 small-size floor. */ @font-face { font-family: 'Arsenal'; font-style: normal; font-weight: 400; font-display: swap; src: url('./fonts/Arsenal/Arsenal-Regular.woff2') format('woff2'), url('./fonts/Arsenal/Arsenal-Regular.ttf') format('truetype'); } @font-face { font-family: 'Arsenal'; font-style: normal; font-weight: 700; font-display: swap; src: url('./fonts/Arsenal/Arsenal-Bold.woff2') format('woff2'), url('./fonts/Arsenal/Arsenal-Bold.ttf') format('truetype'); } @font-face { font-family: 'Arsenal'; font-style: italic; font-weight: 400; font-display: swap; src: url('./fonts/Arsenal/Arsenal-Italic.woff2') format('woff2'), url('./fonts/Arsenal/Arsenal-Italic.ttf') format('truetype'); } @font-face { font-family: 'Arsenal'; font-style: italic; font-weight: 700; font-display: swap; src: url('./fonts/Arsenal/Arsenal-BoldItalic.woff2') format('woff2'), url('./fonts/Arsenal/Arsenal-BoldItalic.ttf') format('truetype'); } /* ═══════════════════ ROLE: IDENTITY · ARABIC — Amiri 1.002 ═══════════════════ [RD3a] A Bulaq Press revival — the أميرية tradition. The Arabic layer carries a real Egyptian typographic lineage without the house ever announcing it. «Not announced, discovered», applied to type. ⚠️ DIGITAL AND SERVICE SURFACES ONLY. Never the object — X12 / Gate 5. unicode-range confines these files to Arabic codepoints, so a Latin-only surface never fetches them. The role stack in tokens-v0_2.css is what routes text here. */ @font-face { font-family: 'Amiri'; font-style: normal; font-weight: 400; font-display: swap; src: url('./fonts/Amiri/Amiri-Regular.woff2') format('woff2'), url('./fonts/Amiri/Amiri-Regular.ttf') format('truetype'); unicode-range: U+0600-06FF, U+0750-077F, U+08A0-08FF, U+FB50-FDFF, U+FE70-FEFF; } @font-face { font-family: 'Amiri'; font-style: normal; font-weight: 700; font-display: swap; src: url('./fonts/Amiri/Amiri-Bold.woff2') format('woff2'), url('./fonts/Amiri/Amiri-Bold.ttf') format('truetype'); unicode-range: U+0600-06FF, U+0750-077F, U+08A0-08FF, U+FB50-FDFF, U+FE70-FEFF; } /* ═══════════════════ ROLE: SUPPORT · ARABIC — Almarai 1.10 ═══════════════════ [RD3b] Four weights cover every role in the support scale, and the support role's job under this house's standard is SILENCE — the system never competes with the object. ⚠️ DIGITAL AND SERVICE SURFACES ONLY. Never the object — X12 / Gate 5. */ @font-face { font-family: 'Almarai'; font-style: normal; font-weight: 300; font-display: swap; src: url('./fonts/Almarai/Almarai-Light.woff2') format('woff2'), url('./fonts/Almarai/Almarai-Light.ttf') format('truetype'); unicode-range: U+0600-06FF, U+0750-077F, U+08A0-08FF, U+FB50-FDFF, U+FE70-FEFF; } @font-face { font-family: 'Almarai'; font-style: normal; font-weight: 400; font-display: swap; src: url('./fonts/Almarai/Almarai-Regular.woff2') format('woff2'), url('./fonts/Almarai/Almarai-Regular.ttf') format('truetype'); unicode-range: U+0600-06FF, U+0750-077F, U+08A0-08FF, U+FB50-FDFF, U+FE70-FEFF; } @font-face { font-family: 'Almarai'; font-style: normal; font-weight: 700; font-display: swap; src: url('./fonts/Almarai/Almarai-Bold.woff2') format('woff2'), url('./fonts/Almarai/Almarai-Bold.ttf') format('truetype'); unicode-range: U+0600-06FF, U+0750-077F, U+08A0-08FF, U+FB50-FDFF, U+FE70-FEFF; } @font-face { font-family: 'Almarai'; font-style: normal; font-weight: 800; font-display: swap; src: url('./fonts/Almarai/Almarai-ExtraBold.woff2') format('woff2'), url('./fonts/Almarai/Almarai-ExtraBold.ttf') format('truetype'); unicode-range: U+0600-06FF, U+0750-077F, U+08A0-08FF, U+FB50-FDFF, U+FE70-FEFF; } /* ═══════════════════ WHAT IS DELIBERATELY ABSENT ═══════════════════ NO @font-face for Alegreya Sans or IBM Plex Sans Arabic, and their files are NOT in this kit — not even now that WOFF2 versions of them exist one folder away. They are pre-authorised fallbacks, invoked THROUGH WASM on a demonstrated hierarchy need — never picked. A face that is present and declared is a face that will get used, so LOCKED has to mean ABSENT, not merely unrecommended in a comment. [RD1 / RD3b] NO Google Fonts , NO @import from any host. If one appears on a house surface, that surface fails the ruling regardless of how it renders. ═══════════════════════════════════════════════════════════════════ */ /* Produced by: Nasaq */