Accessible name and description computation

[danielweck]

https://www.w3.org/TR/accname-1.2/#mapping_additional_nd_te

https://github.com/w3c/accname

[danielweck]

• https://github.com/eps1lon/dom-accessibility-api
• https://github.com/WhatSock/w3c-alternative-text-computation
• https://github.com/AnandChowdhary/get-computed-label

[danielweck]

Thorium issues:

• TTS : aria-labelledby and aria-describedby not read thorium-reader#1933
• TTS readaloud: aria-label on generic elements with overridden ARIA role, still speaks element textual content thorium-reader#1717

[danielweck]

Note role="img" and aria-label="txt" use case:

<p>Roving surveillance underway since November 3, 2008, by the Criminal Intelligence Service of Québec, in the vicinity of the building at <span role="img" aria-label="redacted text">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</span> (former SHELL station) in the municipality of <span role="img" aria-label="redacted text">&#160;&#160;&#160;&#160;</span> has led to the following determinations:</p>

[gautierchomel]

Taking in consideration the following context points:

• TTS is not Screen reader,
• Ebooks are not web sites.

Implementing accessible name should raise us with some attention points:

• Keeping usability without hands on keyboard.
• Not adding to much verbosity and breaks in reading experience.
• Avoiding repetiitions of content reading.

Resuming my understanding of the algorithm cascade proposed by Accessible Name and Description Computation 1.2:

• if hidden is present: read nothing
• Otherwise if aria-labelledby is present: read content of linked element
• Otherwise if aria-describedby is present: read content of linked element
• Otherwise if aria-label is present: read content of aria-label
• Otherwise if title is present: read content of title
• Otherwise if role is present: read content of title

Issues I see:

• aria-labelledby may occur to point to a long element. Should be raised to authoring attention.
• aria-describedby Could be a reference to an entire section, so when reaching the image, the TTS would go reading some paragraphes. It could end up long and will obviously be a repetition as the aria-describedby attribut points to an existing text.
  • Best option would be to prefix and just indicate the portion describing (i.e. Described by "paragraphe 3"). But as authoring exemples are missing and would pobably be completly different from one author to another, I see this as not doable.
  • One option could be to mention something like "press Enter to read description".
  • To my knowledge it is not used by publishers, but could be in academic contexts where there's often a paragraphe describing and commenting the image.
• title is used for tooltips or other purpose (example: epub:type="pagebreak" or role="doc-pagebreak"). Authors might get surprises in earing them loud.
• DPUB specifics roles are missing support in chromium, so won't help a lot.

Proposed behavior:

• Make sure toggle Verbose (or skip) allows readers to lighten they're earing experience
• Ignore aria-describedby in the cascade
• Add DPUB aria role support
• Add user customisation options on strategic elements like epub:type="pagebreak" or role="doc-pagebreak"