- From: John Foliot <john@foliot.ca>
- Date: Tue, 21 Aug 2012 18:56:56 -0700
- To: "'Maciej Stachowiak'" <mjs@apple.com>, "'Silvia Pfeiffer'" <silviapfeiffer1@gmail.com>
- Cc: "'Edward O'Connor'" <eoconnor@apple.com>, <public-html@w3.org>, "'HTML Accessibility Task Force'" <public-html-a11y@w3.org>
Maciej Stachowiak wrote: > > How about just suggesting some sort of notification of the image, > without overconstraining what it is? In general, we let assistive > technologies (or really UA/AT combos) have a lot of freedom in how they > handle repair of broken content. Irrespective of the audio cue being provided, I believe that what is most important is that we communicate via the Accessibility API (and thus most likely via Accessible Name) that this is an image of suspected and likely importance, but that no alternative text is provided. We must remember that not every non-sighted user can actually hear either: how would this manifest via Braille output devices for example? By having a reserved keyword for that Accessible Name I suspect some tools could replace 'word' with 'audio-cue' (I know that Google/ChromeVox are interested in what they dub "ear-cons" - as opposed to eye-cons), but I think leaving that to the screen reader (as opposed to the browser) would be the better choice moving forward. In a Best Case scenario, the 'verbosity' of this reserved keyword could also be controlled by individual screen reading tools (likely via user-settings, such as JAWs Verbosity Settings) and could also be mapped to localization files at the browser level. > For Safari+VoiceOver, saying "Image" > might be a good way to handle these in general. But we might want to > get more fancy - for instance, we could special-case the situation > where the image has a sensible filename (for example "architecture- > diagram.png" or "cat.jpg" instead of "IMG17567.JPG"). And I would not > presume to dictate what is right for other ATs. I'm not too keen on that idea personally, as it begins to presume the reason of/for the image, and may be completely off the mark: for example "cat.jpg" could be a photo of a kitten, a roaring lion, or my friend Catherine... The idea of heuristic guessing for alt text has surfaced previously (it's been a while), and I think that overall this idea is not well liked: you have a better than equal chance of guessing significantly wrong as you do guessing correctly; in actual fact the appropriate alt text for your image might actually be "タイガー" > > BTW, I appreciate that John did research on this question! De nada Maciej. What is important is that we get this (and everything else) right from the end-user perspective. I can appreciate Henri and Mike's perspective on this topic (as developers and maintainers of validation tools), but without a solution that also does something better for the end user, I fear that we are not really solving the real problem. JF
Received on Wednesday, 22 August 2012 01:57:26 UTC