- From: Leif Halvard Silli <xn--mlform-iua@xn--mlform-iua.no>
- Date: Sat, 1 Sep 2012 04:40:35 +0200
- To: Steve Faulkner <faulkner.steve@gmail.com>
- Cc: Mathew Marquis <mat@matmarquis.com>, HTML WG <public-html@w3.org>, "public-respimg@w3.org" <public-respimg@w3.org>
Steve Faulkner, Fri, 31 Aug 2012 20:14:06 +0100: Just a words on what you said complication, Steve: I think it is the fact that <picture></picture> currently is modeled after <img />, with an alt attribute and everything, that complicates. Ultimately, built into the alt attribute idea, lies the img element model or the role=img model: A void element where all the alternative text is provided via attributes. > As previously discussed and bugged I do not support the addition of an alt > The way I envisage <picture> to work is similar to how <canvas> is supposed > to work > So until UAs support picture, text alternatives can be supplied via the > fallback image. It could still be supplied this way for browsers that > support picture btw. > > <picture> > <img alt="text alternative"> > </picture> +1 Perhaps the best thing with the canvas model that the child <img> element becomes an integrated part of the <picture>. Which in turn removes the need - or wish - to duplicate the alternative text in each (fallback) layer. (And thus also removes the need/wisth to get rid of duplication via aria-labelledby ...) Also very good is that the canvas model prohibits that <picture> can be given the ARIA 1.0’s img role as its default role. HTML5 permits adding role=img to the canvas element. But note that doing so immediately causes (at least per the ARIA spec) the content, including the shadow DOM, to become hidden from AT - meaning that authors must provide the alternative text via attributes. The object element does not follow the canvas model - with the object element, then it is an either-or: Either you get the object element, you get its fallback. As a result, to make <object> accessible, one must - at least currently - use aria attributes. (And one probably has to apply role=img to it as well.) Hence I am sceptic about the object model here. The hybrid canvas model seems much more fruitful. But as we discuss adding longdesc to <img>, then I don't quite understand why we should not allow interactive content in <picture>. Perhaps there should be restriction, though - OK. But not forbiddance, me think. And by the way: Would it be authors or AT/UAs that decided the interactivity? E.g. would an <a href> sease to become interactive when inside <picture> or would this be an authoring rule thing? -- Leif Halvard Silli
Received on Saturday, 1 September 2012 02:41:10 UTC