The residual delta is real, it is six criteria wide, and the Typst flag is both stronger and narrower than the parent brief claimed
The question
"Does PDF/UA-1 (ISO 14289) conformance actually reach WCAG 2.2 AA for a static document, or is there a residual delta the Typst ua-1 flag would leave uncovered?"
Context: [[2026-09-03-wcag-gaps-verify-pdf-output]] recommended a three-part fix whose load-bearing first part is flipping typst compile --pdf-standard ua-1, on the stated assumption that PDF/UA-1 conformance closes WCAG 1.1.1, 2.4.2, 3.1.1 and the structure-tree half of 1.3.1. That brief did not verify the mapping. If a delta exists, the recommended split under-covers.
What we already know (from the vault)
- No vault document covers PDF/UA, ISO 14289 or the Matterhorn Protocol. A QMD sweep (lex + vec over
rdco-vault) returns only [[2026-09-03-wcag-gaps-verify-pdf-output]] itself, which raises the question as its own first follow-up. This brief is the first coverage of the standard. - The parent brief's flag claim was verified empirically, not read from docs. It confirmed
--pdf-standard ua-1hard-fails on missing alt text and missing document title on the installed Typst 0.14.2. That part replicates (below). What it did not do is check the converse: what the flag lets through. - The parent brief already routes perception to a separate tier, on the reasoning that "PDF/UA (ISO 14289) tests structure, not perception." That instinct is right in outcome but wrong in one detail, and the detail matters (see Contradictions).
- [[2026-05-20-verify-stack-two-gate-pass-fail-architecture]] gives the tier split this question has to resolve into: a code-based eval tier for arithmetic and a large-language-model-judge tier for judgment. The whole value of answering this question is deciding which residual items go in which tier.
- [[DESIGN-rdco]] is the contract that names Typst as the contract-grade portable-document-format substrate and defers token-pair contrast validation to a "Phase 5," so the Typst toolchain's own limits are binding on any recommendation here.
What the web says
Three layers, kept separate on purpose. Layer (a) is what the standard requires, (b) what the Matterhorn Protocol can mechanically check, (c) what the Typst flag actually enforces.
- (a) ISO 14289-1:2014 is 17 pages and its scope clause excludes nothing about accessibility. Verbatim from the free International Organization for Standardization abstract: the standard "specifies the use of ISO 32000-1:2008 to produce accessible electronic documents," and is "not applicable to" four things, all of them process, rendering implementation, storage media, and hardware. So contrast and reflow are absent from the requirements, not declared out of scope. That distinction is worth keeping; the parent brief's framing implies an explicit exclusion that does not exist.
- (a) The PDF Association states plainly that PDF/UA conformance does not establish content accessibility, and names colour and contrast as the first uncovered case. Verbatim from its official ISO 14289-1 resource page: "By itself, conformity to PDF/UA does not necessarily ensure the accessibility of a document's content. Cases not covered by PDF/UA include, and are not limited to: Authors may have used color or contrast in an inaccessible manner." The same page notes PDF/UA "may be used in conjunction with WCAG 2.x" and that a separate machine-readable PDF Declaration in the Extensible Metadata Platform packet is the mechanism for an actual WCAG conformance claim. Conformance to the one is not a claim about the other.
- (a) The relationship is asymmetric in both directions. Verbatim from the PDF Association's Techniques background page: "A fail case according to PDF/UA-1 may not always be a fail for WCAG, and would therefore be considered as an 'advisory technique' in WCAG terms."
- (b) The Matterhorn Protocol 1.1 is 31 checkpoints and 136 failure conditions, split 87 machine-checkable / 47 human-judgment / 2 with no specific test. Verbatim from page 5. The protocol also says the split is advisory: "While not determinative the value of 'How' generally indicates the realistic best-practice approach at the present time." It explicitly declines to address "partial conformance" and "severity of non-conformance."
- (b) The machine-checkable majority is concentrated in plumbing, not semantics. Checkpoint 31 (Fonts, 30 conditions) and Checkpoint 28 (Annotations, 18) together hold 48 of 136 conditions and roughly half of all machine-checkable ones. Every semantic-judgment area is human-dominated: Tables 1 of 5 machine-checkable, Graphics 1 of 8, Headings 4 of 7, Real-content-vs-artifact 4 of 7, Appropriate Tags 0 of 3. The two conditions that decide whether a cell is a header (15-001, 15-002) are both human. So is "tags are not in logical reading order" (09-001) and the artifact pair "artifact is tagged as real content" / "real content is marked as artifact" (01-001, 01-002).
- (b) Matterhorn has a Color and Contrast checkpoint, and it contains no contrast ratio. Checkpoint 04 has exactly one condition, 04-001, human-only: information conveyed by "contrast, color, format or layout" where "the content is not tagged to reflect all meaning conveyed." That is a WCAG 1.4.1 analogue satisfied by tagging the meaning, not by adding a non-colour cue. There is no 1.4.3, 1.4.6 or 1.4.11 analogue anywhere in the 136 conditions.
- (b) Language of parts is the one place PDF/UA is stronger and more granular than WCAG. Checkpoint 11 has 7 conditions, 6 of them machine-checkable, covering page content, Alt/ActualText/E attributes, outline entries, annotation Contents, form-field TU entries and document metadata. Verbatim from PDF/UA in a Nutshell: "The language of all content must be noted, and changes of language must be explicitly marked as such."
- (W3C) WCAG2ICT settles the applicability questions for a static document, and retires three candidates. Read directly from w3.org/TR/wcag2ict-22/. 1.4.4 Resize Text applies as written, but its added note says viewers "with a 200 percent zoom feature would automatically satisfy this success criterion" unless the content breaks under zoom, so satisfaction is the reader application's, not the document's. 1.4.12 Text Spacing Note 1: it "only applies to non-web documents and software that are implemented using markup languages and allow the user to modify these text spacing properties," which a fixed-layout portable document does not. 1.4.10 Reflow Note 4: "If a non-web document type and its available user agents do not support reflow, it may not be possible for a document of that type to satisfy this success criterion." Meanwhile 2.4.2 is replaced wholesale by a substitute criterion, "2.4.2 Non-web Document Titled," requiring a programmatically determinable Title property, and 2.4.4 Link Purpose applies as written with a note that a link is "any user interface control that behaves like a hypertext link."
Convergences and contradictions
- Convergence on the architecture, not the coverage. The parent brief's three-tier instinct is correct and is independently confirmed by the Matterhorn split: the areas it routed away from the flag (contrast, perception) are precisely the areas with either zero conditions or human-only conditions. The standard's own publisher names colour and contrast as case one of what PDF/UA does not cover. Nothing here overturns the shape of the recommendation.
- Contradiction 1: "PDF/UA tests structure, not perception" is nearly right and wrong in a way that matters. PDF/UA-1 clause 7.1 paragraph 6 does address perception, via Matterhorn 04-001: meaning may not be carried by contrast, colour, format or layout alone. So use-of-colour (WCAG 1.4.1) is in PDF/UA-1's scope. It is just human-only, and the Typst flag implements no check for it. The gap for 1.4.1 is a flag gap, not a standard gap. Getting this backwards would send 1.4.1 to the wrong tier.
- Contradiction 2: language of parts inverts the expected direction, and Typst cannot express it at all. The dispatch expected 3.1.2 to be a PDF/UA gap. It is the opposite: PDF/UA-1 covers it more granularly than WCAG, and 6 of 7 conditions are machine-checkable. The real blocker is one layer down. I compiled
#text(lang: "fr")[je ne sais quoi]inside anendocument under--pdf-standard ua-1: it succeeded, and the output portable document contains exactly one/Lang (en)entry and the stringfrnowhere. Typst 0.14.2 does not propagate a local language change into the structure tree. WCAG2ICT Note 1 offers an escape for "technologies where there is no assistive technology supported method for marking the language" - but the portable-document format supports it fine. The limitation is Typst, so the escape does not apply. - Contradiction 3: the flag is simultaneously stronger and weaker than the parent brief assumed. Stronger: it also hard-fails heading-level skips and empty heading titles, and it refuses
--no-pdf-tagsoutright, none of which the parent credited. Weaker: it passes a document with zero headings anywhere, passes a table with four/TDcells and zero/TH, passes bookmark-suppressed output, and does not require a document date. Four of the five assertions the parent proposed for its new "check 13" are now build-time errors; the fifth is the only one worth keeping.
Synthesis for RDCO
A residual delta exists, it is six criteria wide for a realistic RDCO artifact, and the parent brief's mapping claim holds for the four criteria it named. --pdf-standard ua-1 on Typst 0.14.2 does close 1.1.1 (alt presence), 2.4.2 (document title), 3.1.1 (document language) and the structure-tree half of 1.3.1, and it closes two more the parent did not claim: consecutive heading levels and non-empty heading titles, both inside 1.3.1. I verified the whole boundary by compiling fixtures on the installed 0.14.2 and by extracting the validator's own error-string table out of the binary. So the headline is not "the recommendation was wrong." It is "the recommendation was right about what it checked and silent about two criteria it never considered."
The delta, each line a criterion that applies to a static document and is not guaranteed by the flag:
- 1.4.3 Contrast (Minimum), AA - no requirement anywhere in PDF/UA-1 or the 136 Matterhorn conditions. Verified: grey
#bbbbbbbody text on white, which computes to 1.92:1 against a 4.5:1 minimum, compiles clean under the flag. - 1.4.11 Non-text Contrast, AA - same, no analogue. Directly exposed on the RDCO left-rail and chart-stroke tokens flagged in the parent brief.
- 1.4.1 Use of Color, A - in PDF/UA-1 (clause 7.1-6) but human-only (Matterhorn 04-001) and unimplemented by the flag. Verified: "The red rows failed and the green rows passed" compiles clean.
- 3.1.2 Language of Parts, AA - covered by the standard, mostly machine-checkable, and not emittable by Typst 0.14.2. This is the one hard blocker in the set: it is a toolchain gap, not a process gap, so no check can close it.
- 2.4.4 Link Purpose (In Context), A - no PDF/UA analogue tests whether visible link text is meaningful, and
#linkin Typst 0.14.2 accepts noaltparameter (verified:error: unexpected argument: alt). Typst auto-fills the annotation/Contentswith the raw target URL, which mechanically satisfies PDF/UA-1 clause 7.18.5-2 while doing nothing for 2.4.4. Verified: two "click here" links resolving to different URLs compile clean. - 1.3.1 Info and Relationships, A - residual half. The flag guarantees a structure tree and heading sanity. It does not check table semantics. Verified: a plain
#tableemits four/TDand zero/THand passes. Worse, the flag passes a document with no headings at all, which is Matterhorn 14-001 and human-only.
Three softer residuals belong in the same list at lower weight, because they are genuine but judgment-bound rather than measurable: 1.1.1 adequacy (the flag checks that alt exists, never that it is correct; Matterhorn 13-005, 13-007, 01-001 and 01-002 are all human), 1.3.2 Meaningful Sequence (reading-order correctness, Matterhorn 09-001, human), and 2.4.6 Headings and Labels (descriptiveness). And three candidates the dispatch asked me to test should be retired, not tracked: 1.4.4 Resize Text is the reader application's responsibility per WCAG2ICT; 1.4.12 Text Spacing is scoped out for fixed-layout documents by WCAG2ICT Note 1; 1.4.10 Reflow is enabled by the tagging the flag already guarantees and carries a WCAG2ICT Note 4 escape. Naming them as retired is worth as much as naming the gaps, because the parent brief's pattern of tracking category errors (focus order, keyboard navigation, landmarks) is the failure mode to avoid repeating.
Does the three-part recommendation under-cover? Yes, in two places, and it over-specifies in a third. It under-covers by omitting 2.4.4 and 3.1.2 entirely: no tier touches either, and 3.1.2 cannot be fixed by a check at all. It over-specifies "check 13," which proposed asserting /StructTreeRoot present, /Lang set, /Figure count equal to /Alt count, no heading-level skip, and tables carrying /Table and /TH. Four of those five are now build-time hard errors, so putting them in a large-language-model critic is duplicated spend - exactly the argument the parent brief itself made against adding critic surface. Only /TH survives. The token-pair contrast validator in part two is necessary and is still the single best item in the recommendation, because it fixes the defects at contract level where they propagate to web, video and application surfaces. But it is not sufficient for 1.4.3: my failing fixture used an ad-hoc rgb("#bbbbbb") that appears in no DESIGN.md token table and would pass a token validator while rendering at 1.92:1. Palette validation and artifact validation are different claims.
The minimal addition is four mechanical checks, one source lint, and one honest expansion of the human gate. Mechanically automatable, zero large-language-model cost, belongs in the code-eval tier: (13) table semantics - every /Table subtree contains at least one /TH, and every /TH carries a /Scope (Matterhorn 15-003, the one machine-checkable table condition); (14) heading presence - heading-tag count greater than zero, and the document outline is non-empty; (15) link text - no link's visible text matches a deny-list (click here, read more, here, learn more, this link), and no two links with identical visible text resolve to different target URIs, both pure string comparisons over the extracted text and annotation dictionaries; (16) language-of-parts source lint - fail the build on any text(lang:) in the .typ source that differs from the document language, with the message that 0.14.2 cannot represent it. Check 16 is a blocker, not a validator: it stops us shipping a 3.1.2 failure we cannot fix, and it should carry a pointer to the Typst issue tracker rather than a remediation step. Semi-mechanical and worth building but worth distrusting: a rendered contrast sampler over the page raster, reliable for flat backgrounds and unreliable over photographs and gradients, which is why it cannot replace the token validator. And the human gate must stay, and must grow by three items: alt-text adequacy, decorative-versus-real classification, reading-order correctness, heading descriptiveness, use-of-colour, images-of-text judgment, and whether a declared language is appropriate (Matterhorn 11-007). Every one of those maps to a Matterhorn condition tagged human-only. That is not a scoping failure on our part; it is the protocol's own assessment of its automation ceiling.
Calibration. The mapping is mostly sound with a real but bounded delta, and I want to be clear that this research did not overturn the parent brief. Confidence is high on layer (c): I ran the compiler on the installed 0.14.2 in both passing and failing configurations for every claim above, and cross-checked against the validator's error-string table lifted from the binary. High on layer (b): the 87/47/2 split and every per-condition machine/human tag are verbatim from the Matterhorn Protocol 1.1 document. High on the WCAG2ICT applicability reads, which came from the W3C note directly. Medium-high, and this is the one real weakness, on layer (a): I read ISO 14289-1's scope clause verbatim from the International Organization for Standardization's free abstract page, but I did not read the normative body of the standard. Every claim about what clause 7 requires is the PDF Association's restatement - the Matterhorn conditions' UA1:7.x citations plus PDF/UA in a Nutshell - not the standard's own words. That is a cheap gap to close and I did not close it: the PDF Association distributes ISO 14289-1 at no cost, sponsored, 17 pages, but behind a cart and checkout flow that a research sub-agent should not complete on the founder's behalf. Worth one minute of his time if any of this becomes a published claim rather than an internal build spec.
Why this is in the vault
This closes the first and most load-bearing open follow-up in [[2026-09-03-wcag-gaps-verify-pdf-output]], whose part-one recommendation (flip --pdf-standard ua-1) is still unbuilt and was resting on an unverified standards mapping. It converts that recommendation from "flag plus validator equals AA" into a specific, costed amendment: keep part one, keep part two, replace the proposed check 13 with four narrow mechanical checks plus a source lint, and accept that 3.1.2 is blocked at the Typst layer rather than closable by a check. It also corrects two inversions that would have mis-tiered the work (1.4.1 is a flag gap not a standard gap; 3.1.2 is a toolchain gap not a standard gap).
Open follow-ups
- Does the PDF Declaration mechanism change what RDCO can honestly claim? The PDF Association describes a machine-readable WCAG conformance claim written into the Extensible Metadata Platform packet, separate from the PDF/UA identifier. Whether Typst can emit one, what the declaration schema requires an author to attest, and whether making an unverified claim is worse than making none are all document-answerable and bear directly on whether RDCO artifacts should carry any accessibility badge.
- Has PDF/UA-2 (ISO 14289-2:2024) added any contrast or link-purpose requirement, and is there a Matterhorn successor for it? Matterhorn 1.1 states on its face that it does not apply to PDF 2.0 or PDF/UA-2. If the second part narrows the delta, the tier split above is sized against a superseded target. Typst only offers
ua-1today, so this is a forward-looking question about whether to wait. - Read ISO 14289-1's normative clause 7 directly via the PDF Association's zero-cost channel, and check whether the "conformity to PDF/UA does not necessarily ensure accessibility" sentence is the standard's own Introduction text or PDF Association editorial. This is the single labelled weakness in this brief's confidence, it is 17 pages, and it is free.
Related
- [[2026-09-03-wcag-gaps-verify-pdf-output]]
- [[2026-05-20-verify-stack-two-gate-pass-fail-architecture]]
- [[2026-07-03-verify-skills-pass-disagreement-audit]]
- [[DESIGN-rdco]]
- [[DESIGN-mac]]
- [[DESIGN-squarely]]
- [[DESIGN-sanity-check]]
Sources
Research trail - what was reachable and what was not. iso.org and all of pdfa.org return HTTP 403 to both curl and WebFetch (Cloudflare bot block); everything from those two hosts was retrieved through headless Chrome instead, consistent with the standing note that Playwright passes Cloudflare blocks. The two PDF Association documents were read via in-page pdf.js rather than downloaded. One paywall was encountered and skipped, not circumvented: ISO's own CHF 100 store listing for ISO 14289-1. A zero-cost sponsored channel for the same document exists at the PDF Association but routes through a cart and checkout, which was deliberately not completed; this is the only unreachable primary source and it is the cause of the single medium-high confidence label in the Synthesis. Two URLs in the original dispatch were stale: pdfa.org/resource/matterhorn-protocol/ 404s (the slug is now the-matterhorn-protocol), and the PDF Association's browsable Matterhorn database at taggedpdfschool.pdfa.org/matterhorn has confirmed fidelity defects against the PDF (checkpoint labels 26/27/28 are shifted, condition 13-008 is missing, and the two no-test conditions are mislabelled as human), so the PDF was treated as authoritative for every quoted condition.
Local primary - measured, not read:
- Toolchain:
typst 0.14.2at/opt/homebrew/bin/typst.--pdf-standardacceptsua-1among 17 values. - Validator error-string table extracted from the binary via
strings -n 10, yielding the full set of PDF/UA and PDF/A conditions krilla implements: missing document language, missing document title, missing alt text, missing heading title, missing document outline, missing annotation alt text, missing document date, missing document tags, image smooth scaling, plus PDF/A-only transparency, font-licence and codepoint conditions. - Fixtures compiled under
--pdf-standard ua-1in/tmp/uatest/. Hard-failed: heading level 1 to 3 skip ("heading levels must be consecutive"); image with noalt; empty heading title; absent#set document(title:);--no-pdf-tagscombined withua-1("cannot disable PDF tags when exporting a PDF/UA-1 document"). Compiled clean:#bbbbbbbody text on white; colour-only meaning ("the red rows failed and the green rows passed"); twoclick herelinks to different URLs;#text(lang: "fr")inside anendocument; a#tablewith notable.header; a document with zero headings; all headingsbookmarked: false; no#set document(date:). - Structure verified with
pdfinfoplus/S /Typescans of the uncompressed structure tree: plain#tableyields 4/TDand 0/TH;table.header(..)yields 2/THand 2/TD, so the fix is expressible.#text(lang: "fr")yields one/Lang (en)and nofranywhere, so that fix is not.#linkrejects analtargument; its/Contentsis auto-filled with the raw target URL. - Typst does expose
pdf.artifact,pdf.header-cell,pdf.data-cellandpdf.table-summary, so the table and artifact remediations are authorable in 0.14.2 even though the flag does not require them. - Contrast ratios recomputed with the WCAG relative-luminance formula:
#bbbbbbon white = 1.92:1; and re-confirming the parent brief,#5a6e82on#eaf2f8= 4.65:1,#F4F1EAon#FF3E2F= 3.11:1,#7A7A78on#F4F1EA= 3.81:1. - Vault gate: QMD lex + vec over
rdco-vaultfor PDF/UA, ISO 14289, Matterhorn and the Typst flag returns no prior coverage.
Web primary:
- https://www.iso.org/standard/64599.html - ISO 14289-1:2014 abstract and scope, read verbatim; 17 pages, confirmed current in 2025
- https://isme.me/en/project/show/iso:proj:64599 - independent free mirror corroborating the scope wording verbatim
- https://pdfa.org/resource/iso-14289-pdfua/ - "cases not covered by PDF/UA," the WCAG relationship statement, the PDF Declaration mechanism, and the zero-cost download offer
- https://pdfa.org/techniques-for-accessible-pdf-background/ - the asymmetry statement between PDF/UA-1 failure cases and WCAG success criteria
- https://pdfa.org/wp-content/uploads/2021/04/Matterhorn-Protocol-1-1.pdf - Matterhorn Protocol 1.1, 20pp, CC BY 4.0; primary source for the 87/47/2 split and every per-condition machine/human tag
- https://pdfa.org/wp-content/uploads/2022/05/PDFUA-in-a-Nutshell.pdf - language-of-all-content requirement, the visual-means-alone prohibition, assistive-technology obligations
- https://pdfa.org/iso-14289-2-pdfua-2/ - PDF/UA-2 parallel "context and limitations" wording, noted but not relied on
- https://www.w3.org/TR/wcag2ict-22/ - WCAG2ICT, read directly for non-web-document applicability of 1.1.1, 1.3.1, 1.4.1, 1.4.3, 1.4.4, 1.4.5, 1.4.10, 1.4.11, 1.4.12, 2.4.2, 2.4.4, 3.1.1, 3.1.2
- https://pdfa.org/infographics-pdfua-and-wcag-2-0/ - the PDF Association's 2015 WCAG 2.0 coverage infographics, surfaced via search and used only as corroboration; predates WCAG 2.1 and 2.2, so treated as secondary
Paywalled, skipped: ISO's own CHF 100 listing for ISO 14289-1:2014. Unreachable: the normative body text of ISO 14289-1:2014 (clause 7 and Introduction), free but behind a checkout flow.