@surea11y/pack-raweb 0.0.0-stage → 1.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,145 @@
1
+ # Rule map review: RAWeb 1.1 against pack-rgaa's links
2
+
3
+ pack-raweb takes `@surea11y/pack-rgaa`'s rule map for RGAA 4.1.2 and keeps, changes or drops each link where RAWeb 1.1 says something different. [`RGAA_COMPARISON.md`](./RGAA_COMPARISON.md) shows every difference. This document records the judgement in two parts:
4
+
5
+ 1. The 47 links on the 29 tests RAWeb rewords.
6
+ 2. The links on tests whose wording RAWeb keeps but whose methodology, particular cases, technical notes or glossary it changes.
7
+ 3. The 8 tests RAWeb adds under RGAA's criteria.
8
+
9
+ The other links carry over as they are.
10
+
11
+ ## How the decisions were checked
12
+
13
+ - **RAWeb's text:** each decision rests on RAWeb's wording, and on its [revision notes](https://github.com/accessibility-luxembourg/ReferentielAccessibiliteWeb/blob/9ebf98f4e1239ccd99dc95d52c81903562075f78/en/notes-de-revision.md) (commit `9ebf98f`), which say why each criterion changed. There is no other reference: RAWeb's methodology is a manual audit.
14
+ - **The rules:** what each rule does is read from its source and its row note in pack-rgaa 1.1.
15
+ - **Test pages:** each change was run on a small page built for it. Each one confirmed that the link, as pack-rgaa makes it, would fail a page RAWeb accepts. These pages become tests of the standard.
16
+ - **The principle:** a link that could fail a page RAWeb accepts is dropped or made a question. A question is never a false failure.
17
+
18
+ ## Summary
19
+
20
+ | | Links |
21
+ | ----------------------------------------------------------------- | -----: |
22
+ | **Part 1: reworded tests** | **47** |
23
+ | Kept as they are | 34 |
24
+ | Kept for some findings, or as questions | 4 |
25
+ | Dropped | 9 |
26
+ | **Part 2: tests with the same wording** | |
27
+ | Made questions where RAWeb now allows what they fail | 2 |
28
+ | Dropped | 1 |
29
+ | **Part 3: tests RAWeb adds** | |
30
+ | Moved to a new test | 1 |
31
+ | Linked to a new test | 3 |
32
+ | **Added** to reworded tests: rules that check what RAWeb now asks | 3 |
33
+
34
+ # Part 1: reworded tests
35
+
36
+ ## Kept as they are (34)
37
+
38
+ | Test | Rules | Why the change doesn't affect them |
39
+ | -------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
40
+ | 1.2.1 | `image-redundant-alt`, `rgaa-img-no-alt-decorative` | RAWeb adds a third way to mark a decorative `<img>`: `role="presentation"` or `"none"` with no `title`, `aria-label` or `aria-labelledby`. The first rule only asks. The second already leaves out images with those roles, so it fails only an `<img>` none of the three conditions accept. |
41
+ | 1.2.4 | `rgaa-svg-decorative-aria-hidden` | RAWeb keeps the one condition this rule checks: a decorative `<svg>` has `aria-hidden="true"`. |
42
+ | 1.3.1 | `image-redundant-alt`, `img-alt-quality`, `rgaa-img-decorative-no-alternative` | Only "pour" ("for") is added at the start. |
43
+ | 1.4.4 to 1.4.7 | `rgaa-captcha-review` | RAWeb removes `alt` from objects, embeds, SVG and canvas, which have no such attribute. The rule only asks. |
44
+ | 4.4.1 | `video-caption` | RAWeb says what relevant captions are (in the video's language, relevant, synchronised) and asks it of at least one track. The rule asks whether the captions are relevant, and that question still holds. |
45
+ | 8.1.1 | `rgaa-doctype-present` | "Is each page defined by a document type?" A page with no doctype isn't. |
46
+ | 8.2.1 | `duplicate-id`, `rgaa-aria-attribute-conformance`, `rgaa-aria-role-conformance`, `rgaa-html-elements-attributes-valid`, `rgaa-html-nesting-valid`, `rgaa-markup-validation-review` | RAWeb asks the whole question, "is the generated source valid for the document type?", in place of RGAA's list of conditions. Each rule fails a defect that makes the source invalid. RAWeb relates 8.1 and 8.2 to WCAG 4.1.1, so its profile keeps 4.1.1 (`keepsParsing`), as pack-rgaa's does. |
47
+ | 10.7.1 | `css-focus-indicator-suppressed`, `css-hidden-focus`, `manual-review`, `rgaa-focus-indicator-contrast` | "Du clavier" ("keyboard") is added: the revision notes say the test concerns keyboard focus, not mouse focus. The rules already check keyboard focus. |
48
+ | 10.12.1 | `avoid-inline-spacing`, `text-spacing-content-loss` | "Increased to" becomes "set to at least". The values are WCAG 1.4.12's, which both rules apply. |
49
+ | 10.13.2 | `rgaa-tooltip-review` | An accent only. |
50
+ | 11.1.3 | `form-control-label-quality`, `form-control-programmatic-label-quality`, `rgaa-field-hidden-label-review` | A typo only. |
51
+ | 11.10.4 | `invalid-field-error-message` | RAWeb extends the test from required fields to every field with `aria-invalid="true"`. The rule asks about every such field, so the link fits better than before. |
52
+ | 13.5.1 | `rgaa-cryptic-content-review` | RAWeb no longer accepts a `title`; the meaning must be in the adjacent context or given in a way assistive technologies support. The rule asks about cryptic content that has no `title`. Content that has one isn't asked about, which is a gap but never a false failure (see the gaps below). |
53
+ | 13.8.1, 13.8.2 | `rgaa-moving-content-review` | RAWeb adds "excluding special cases": essential movement, and loaders. The rule only asks. |
54
+ | 13.9.1 | `css-orientation-lock`, `rgaa-orientation-content-parity` | Grammar only ("peut" becomes "peuvent"). |
55
+
56
+ ## Kept for some findings, or as questions (4)
57
+
58
+ | Test | Rule | Decision | Why |
59
+ | ------- | ------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
60
+ | 1.2.1 | `rgaa-img-decorative-no-alternative` | `hiddenImgHasAlternative` (an `<img aria-hidden="true">` with a `title` or ARIA name) becomes a question under 1.1.1, in place of a failure of 1.2.1. | Condition 2 accepts `aria-hidden="true"` alone, and the revision notes say it is enough for the image to be ignored. If the image is decorative, it passes 1.2.1. If it carries information, hiding it fails 1.1.1, which only a person can tell. The other code, `decorativeImgHasAlternative` (`alt=""` or a presentation role with a name), is a question under 1.2.1, 1.1.1 and 1.3.1, as before: condition 1 and the new condition 3 both forbid the name. |
61
+ | 9.2.1 | `rgaa-main-element-structure` | Only `noMain`, `mainMayBeInFrame` and `allHidden` count. `roleMainOnly`, `roleZoneOnly`, `extraMain`, `revealableMain` and `hiddenByAncestor` count toward no test. | RAWeb asks for a computed role (`banner`, `navigation`, `search`, `main`, `contentinfo`), given by the HTML element or by `role`. So `role="main"` on a `<div>` passes, and so do the other zone roles on any element. RAWeb also dropped the single visible `<main>` condition. A second `<main>` without `hidden` is still invalid HTML, and `rgaa-html-nesting-valid` fails it under 8.2.1. |
62
+ | 11.10.1 | `rgaa-required-field-indication` | Every failure becomes a question, under 11.10.1 and 11.10.2. `requiredNotIndicated`, `allFieldsRequiredSentence` and `markerNotExplained` count toward 11.10.2 only; `markerExplainedAfter` and `signMayMeanOther` toward both. | RAWeb changes 11.10's particular cases. A form with only a username and a password field is now outside 11.10.1 and 11.10.2. A form that marks its optional fields no longer is. The rule doesn't recognise a login form, so on one it fails the required fields RAWeb leaves out; the test page confirmed it. Only a person can tell a login form, so its failures become questions. Separately, those three codes are given only for fields with `required` or `aria-required="true"`, and condition 2 of 11.10.1 accepts such a field. RGAA 4.1.2 has the same condition, and pack-rgaa 1.1 counted two of them (`requiredNotIndicated`, `markerNotExplained`) toward 11.10.1 too. The criterion's verdict was right either way, but the test it was attributed to was wrong. pack-rgaa's next release fixes it. |
63
+ | 11.10.2 | `rgaa-required-field-indication` | Every failure becomes a question (see 11.10.1 above). | RAWeb adds two ways to pass: a visible indication that names the field, and one in the group's legend. The rule fails a field only when its label, linked passages and surroundings show no indication, and asks when one is in the legend or next to the field. A general sentence such as "all fields are required" names no field. The failures still become questions, for the login-form particular case. |
64
+
65
+ ## Dropped (9)
66
+
67
+ | Test | Rule | Why |
68
+ | -------------- | --------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
69
+ | 1.2.4 | `rgaa-svg-hidden-no-alternative` | RAWeb removes conditions 2 to 4 (no text alternative, no `<title>` or `<desc>`, no `title`): `aria-hidden="true"` is enough, as the revision notes say. An `<svg aria-hidden="true">` with a name passes 1.2.4. The rule's question under 1.1.5 (is the image informative, and so wrongly hidden?) stays. |
70
+ | 1.2.4 | `svg-text-alternative-quality` | For the same reason, an `<svg>` that turns out to be decorative and keeps its alternative no longer fails 1.2.4. The question stays under 1.3.6. |
71
+ | 1.9.1 to 1.9.5 | `rgaa-figure-caption-structure` | RAWeb removes the conditions this rule checks: `role="figure"` or `"group"` on `<figure>`, and an `aria-label` matching the caption. The conditions RAWeb keeps (the image and caption in a `<figure>`, the caption in a `<figcaption>`) are what make the rule apply, so it checks nothing RAWeb asks. |
72
+ | 6.2.1 | `rgaa-link-content-label-present` | RAWeb 6.2.1 asks only for an accessible name. Its revision notes and glossary drop RGAA's requirement for content between `<a>` and `</a>`, and the note that discounted text generated by CSS. A link named by `aria-label`, `aria-labelledby` or `title` passes, and this rule fails it. |
73
+ | 12.6.1 | `rgaa-page-zones-reachable` | RAWeb 12.6.1 now excludes the main zones (header, navigation, main content, footer, search), which move to 9.2.1. Those zones are exactly what this rule looks for, and its questions follow RGAA's 12.6 conditions. |
74
+
75
+ ## Added (3)
76
+
77
+ | Test | Rule | Why |
78
+ | ----- | --------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
79
+ | 6.2.1 | `link-name-present` (core) | It fails a link with no accessible name at all, which is exactly what RAWeb 6.2.1 now asks. pack-rgaa leaves it unlinked because RGAA asked for more. |
80
+ | 8.1.1 | `rgaa-doctype-valid`, `rgaa-doctype-position` | RAWeb asks whether each page is defined by a document type, in place of RGAA's three tests (present, valid, before `<html>`). A doctype named other than `html`, or written after `<html>` (which the parser drops), doesn't define the page for user agents, and that is how the glossary "Type de document" ("document type") describes its purpose. `rgaa-doctype-valid`'s question about an unknown public identifier stays a question. |
81
+
82
+ # Part 2: tests with the same wording
83
+
84
+ RAWeb changes the methodology of 43 tests that have rules, the particular cases or technical notes of 20 criteria, and 30 glossary entries ([`RGAA_COMPARISON.md`](./RGAA_COMPARISON.md)). Most changes are editorial: wording, Flash and Internet Explorer notes removed, French references replaced by Luxembourg's. The ones below change what a rule may report.
85
+
86
+ ## Made questions where RAWeb now allows what they fail (2)
87
+
88
+ | Test | Rule | Decision | Why |
89
+ | ------------ | --------------------------------------- | ---------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
90
+ | 10.1.2 | `rgaa-presentational-attributes-absent` | A finding on an `<iframe>` or a `<source>` becomes a question. Every other finding still fails. | RAWeb's methodology allows `width` and `height` on `<iframe>` and on any `<source>`, where RGAA allows neither. So the rule would fail every embedded player that sets its size. RAWeb still forbids `frameborder` or `align` on those elements, and the rule reports all of an element's attributes in one finding. So the finding becomes a question rather than being dropped. |
91
+ | 5.7.2, 5.7.3 | `rgaa-th-scope-row-col` | `scope="rowgroup"` or `"colgroup"` becomes a question under 5.7.2, 5.7.3 and 5.7.6. Any other value still fails. | RAWeb adds test 5.7.6, which accepts `scope="rowgroup"` and `"colgroup"` on a header over a group of rows or columns. RGAA accepts only `row` and `col`. Whether a header covers a group is for a person: if it does, it passes 5.7.6; if it doesn't, it fails 5.7.2 or 5.7.3. |
92
+
93
+ ## Dropped (1)
94
+
95
+ | Test | Rule | Why |
96
+ | ------------ | ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
97
+ | 6.1.2, 6.1.3 | `rgaa-link-title-content-name` | The rule fails an image or composite link whose `title` doesn't repeat the alternative of its images. Its only ground is note 2 of RGAA's glossary entry "Intitulé (ou nom accessible) de lien" (link name), which RAWeb removes, with notes 1 and 3 to 5. Under RAWeb's name order (`aria-labelledby`, `aria-label`, content, `title`), the content names the link and the `title` doesn't. Where a link's visible text must be in its `title`, 6.1.5 asks it, and `rgaa-link-label-in-name-sources` checks that. |
98
+
99
+ ## Kept, and why the change doesn't affect them
100
+
101
+ - **Tables, 5.1.1 and 5.2.1:** RAWeb words the complex-table definition differently: headers "over several rows and/or columns" rather than "not only in the first row and/or column". The rules only ask whether a table is complex.
102
+ - **Redirects, 13.1.2:** RAWeb's methodology accepts a redirect after 20 hours or more, where RGAA's doesn't. `rgaa-meta-redirect-immediate` only asks about a delayed redirect.
103
+ - **Moving content, 13.8:** the glossary now says pausing on hover isn't control. The rule only asks.
104
+ - **Players, 4.7.1:** the methodology drops Flash and the condition that the identifying text sits outside the player. The rule only asks.
105
+ - **Documents, 13.3.1:** a PDF is now judged against RAWeb's PDF framework (RAPDF). The rule only lists the documents.
106
+ - **SVG, 1.1.5:** RAWeb adds the `<title>` element to the sources of an alternative. `rgaa-svg-role-img` and `svg-text-alternative-present` already count it.
107
+
108
+ # Part 3: tests RAWeb adds
109
+
110
+ RAWeb adds 8 tests under RGAA's criteria: 4.3.3 and 4.4.2 (live media), 5.7.6 (group headers), 6.1.6 (identical links), 9.2.2 to 9.2.4 (the document structure) and 13.5.2 (cryptic content hidden from assistive technologies). A rule is linked to one only where it checks what the test asks.
111
+
112
+ ## Moved to a new test (1)
113
+
114
+ | Test | Rule | Why |
115
+ | ----- | ------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
116
+ | 6.1.6 | `identical-links-same-purpose` | It asks whether links with the same name and different destinations serve the same purpose, which is exactly 6.1.6: "Does each identical link have the same function and destination?" Under RGAA it was linked to 6.1.1 to 6.1.4 through note 3 of the glossary entry for a link's name, which RAWeb removes; RAWeb's new glossary entry "Lien identique" (identical link) takes its place. Core files the rule under WCAG 2.4.9, which RAWeb doesn't relate to criterion 6.1; the row says why it is linked all the same. |
117
+
118
+ ## Linked to a new test (3)
119
+
120
+ | Test | Rule | Why |
121
+ | ----- | ------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
122
+ | 9.2.2 | `landmark-no-duplicate-main`, `landmark-no-duplicate-banner`, `landmark-no-duplicate-contentinfo` | 9.2.2 asks for a single visible element with each of the computed roles `main`, `banner` and `contentinfo` (conditions 1 to 3). Each rule asks about a page exposing more than one landmark of its role. They count only landmarks exposed to assistive technologies, and only ask. |
123
+
124
+ 5.7.6 is under Part 2, with `rgaa-th-scope-row-col`. No rule checks 4.3.3, 4.4.2, 9.2.3, 9.2.4 or 13.5.2.
125
+
126
+ ## Two tests without RAWeb's methodology
127
+
128
+ RAWeb 1.1 publishes no methodology for 8.1.1 and 8.2.1, in either language. A person still needs steps, so the pack gives each two of its own, drawn from the test's question and the glossary entry "Type de document" (document type). The criterion's note says they are the pack's.
129
+
130
+ ## Gaps, and how two of them were filled
131
+
132
+ Two gaps found by this review were filled with rules of the packs this one builds on, rather than of its own:
133
+
134
+ - **9.2.1, 9.2.3 and 9.2.4:** core's `page-zones-landmark` (core 1.17) asks about a block named as the page header, a navigation or the footer, or a search form, that has no computed role of its nature, and about a page header or footer nested in a section, where it loses its role. Its codes are linked to 9.2.1, and `HEADER_NESTED` and `FOOTER_NESTED` to 9.2.3 and 9.2.4. It only asks.
135
+ - **13.5.1:** pack-rgaa's `rgaa-cryptic-content-review` reports cryptic content with a `title` under a code of its own, `titleOnly`, which RGAA counts toward no test and RAWeb links to 13.5.1 as a question.
136
+
137
+ What no rule checks yet:
138
+
139
+ - **9.2.2's fourth condition:** navigation landmarks used only for navigation zones.
140
+ - **9.2.1 on a page whose doctype isn't HTML5:** `rgaa-main-element-structure` doesn't apply there, though RAWeb removed that special case. The result is no finding, never a false one.
141
+ - **12.6.1:** no rule. Which blocks are grouping zones present on several pages is a person's call.
142
+ - **1.9.1 to 1.9.5:** no rule checks that a captioned image and its caption sit in a `<figure>` with a `<figcaption>`: which text is an image's caption can't be told reliably.
143
+ - **13.5.2:** cryptic content hidden from assistive technologies.
144
+ - **11.10.1 and 11.10.2:** `rgaa-required-field-indication` leaves out a form that marks its optional fields, which RGAA made a particular case and RAWeb doesn't. It also looks only at a `<form>` or `role="form"`, where RAWeb's glossary entry "Formulaire" (form) adds that a form needn't be in one.
145
+ - **5.1.1:** a `summary` attribute before HTML5, which RAWeb no longer accepts, isn't asked about.