- From: Sander Tekelenburg <st@isoc.nl>
- Date: Mon, 6 Aug 2007 09:01:51 +0200
- To: public-html@w3.org
At 11:36 +1000 UTC, on 2007-08-06, Jason White wrote: > On Sun, Aug 05, 2007 at 03:42:45PM +0200, Sander Tekelenburg wrote: [limiting <alt> to refer to specific elements] > If there are use cases for providing alternatives to other > elements, then I have no objection to allowing its application to them as > well. There might be good reasons to have to limit <alt> to specific elements. I haven't yet thought about that muchy. But I think that limits in general should only be defined when they're necessary. So my initial take would be to allow <alt> to refer to any element. If we limit <alt> to referring to only specific elements, then we might end up making authors use home-grown workarounds (likely to generate problems) for situations that we didn't anticipate. That's just my principle way of looking at things in general. But it also seems in fact likely to me that authors would in fact have a use to refer <alt> to other elements. Two examples: [1] An author might feel that the ideal way to convey something is through a table, but realise that it's also possible to express the same through prose and realise that that would be easier to digest for some users. (Because he knows that for most users the table will convey the message 'in a blink', he considers the <table> primary, and doesn't want to bother those users by also confronting them with the prose version.) [2] An author may at times feel that prose is the best way to explain something, but realise that for some users an image, audio file, video might be eaiser to digest. (For example because he feels that the prose can express his message the most precise, but estimates that that precision will make it too hard to digest for some users. An image, audio or video might be easier to digest for those users.) So unless there are reasons to have to forbid, it seems wise to me to allow it all -- to only require that <alt> refers to an element within <body>. > However, if the intent is to restrict it to images and media elements, > then <alt for=idref> where idref refers to another type of element, should > result in a validation error, as it is then a mistake on the part of the > author. Yes. But note that for HTML5 that would not be enough. You'd also need to spec how UAs would have to treat that error. [allow author to define fallback cascade] >> > <embed [...] id="a"> >> > <alt for="a" id="b"> [...] >> > <alt for="b" id="c"> [...] [...] > Nested ALT elements are another possibility, similar in semantics to nested > <object>, again assuming that a full hierarchy is desired. This would be >less > prone to error on the author's part, I expect. Agreed. That'd likely be less prone to errors. And UA how to consider accidentally omitted closing tags and such is already defined. Implementation aside, I'm not sure though that authors should suggest (or even define) a falback cascade order. Are they in a position to know which order would be best? [...] <aside> >> I see @alt and @longdesc as a single >> two-piece hack to overcome the unique problems of <img>. Each serves one >>half >> of the same goal. [...] >> >> Coming from that view, I'd think that the content of <alt> would replace >>both >> @alt and @longdesc. With <alt>, there is no longer a need to think in terms >> of <img>'s split fallback personality, and I think that would be very >> healthy. [...] > > I agree, though I would argue that where @alt and 2longdesc both exist in >HTML > 4, @alt is effectively providing advisory information about the resource > referred to by @longdesc. I don't think of it like that at all :) In HTML 4, @alt is to provide an equivalent. Period. If any of @alt and @longdesc are about advisory information, it would be @longdesc. (Consider Gregory's example of a piicture of a flag to point to another language version of a document, with @alt saying "german", and @longdesc describing the german flag. (Although I'm personally not convinced that @longdesc was ever intended for that or should be used for that.)) (Anyway, this is perhaps not that essential for what we're discussing.) </aside> > In practical terms, if an author has written multiple paragraphs of > explanation as the content of <alt>, there is considerable value, from the > user's point of view, in having a short label or summary in the TITLE > attribute of <alt> so that one can decide whether and when to read the > detailed explanation. Absolutely, yes. > Although this may be characterized as advisory > information, it would require something better than title="description of > image" or title="transcript of audio". It seems to me that the title attribute to <link> is comparable here: <link rel="next" title="next page"> <link rel="section" title="products"> <link rel="section" title="documentation"> <link rel="help" title="support"> <link rel="stylesheet" title="default look"> <link rel="alternate stylesheet" title="fancy look"> etc. The reason that such short advisory notes work (and are probably even preferable), is that the information is accessed throgh the chrome and thus in a manner that is consistent across sites. In other words the user is already provided with an easily reconisable context. [...] -- Sander Tekelenburg The Web Repair Initiative: <http://webrepair.org/>
Received on Monday, 6 August 2007 07:03:55 UTC