Re: unifying alternate content across embedded content element types

Hi Jon,

> On Jul 15, 2007, at 9:06 AM, Jon Barnett wrote:
>
> On 7/15/07, Robert Burns <rob@robburns.com> wrote:
>>
>>  My contention (and this was a point I think Jon was making too), is
>>  that when a textual equivalent describing the image is already in  
>> the> > surrounding  prose, it shouldn't be repeated for the @alt  
>> attribute.
>
> Yes, that's a part of what I was saying.
>
>>  My proposal was something shorter that let's the user know what
>>  meaning the image conveys in a way that matches it with the prose
>>  available to all users.
>
> Sorry for my ambiguous use of the word "alternate".  By "equivalent"
> != "alternate", I meant that "equivalent" is a subset of "alternate".
>
> Sometimes alternate content is used as a drop-in equivalent:
> Completely replacing the image with the equivalent content doesn't
> change the meaning of the document
>
> Sometimes alternate content is used as a description of the media:
> Complete replacing the image with the descriptive (alternate) content
> doesn't make sense unless you let the user know that something is
> missing.

I think it is useful to let users know when they are viewing fallback  
(in terms of the authors intent it's fallback). However that is true  
at all times (whether descriptive or equivalent to use your words).  
My contention is that none of the fallback should be merely  
descriptive and not equivalent (again in the manner that I understand  
you using these terms). (Though I guess when both @alt and @longdesc  
are attached to an element, it might be sufficient to include less  
than is needed to understand the document in the @alt as long as its  
supplemented by the @longdesc target.) Am I using your terms correctly?

> And my proposal is that the same markup used to provide equivalent
> (alternate) content shouldn't be the same as the markup used to
> descriptive (alternate) content.
>
> With relation to this thread, I meant that I don't like <object alt>
> as a mechanism for "short alternative" content.  Would it be an
> equivalent to the <object>, or would it be a description of the
> object?  I think it has too much potential for misuse in the same way
> <img alt> is being misused.  (And the fact that professionals can
> disagree on <img alt> tell me that it's not defined well enough)

I'm much less concerned that @alt is being used improperly than I am  
concerned that it isn't being used at all or isn't even available on  
other elements.

> <p>A paragraph about Fluffy's kittenhood...
> <img src="fluffy-yarn.jpg" alt="">
> <p>On Saturday, Fluffy was playing with a ball of red yarn in my
> living room.  It was quite funny.
>
> In this context @alt isn't necessary.

Agreed.

>
> <p>A paragraph about Fluffy's kittenhood...
> <img src="fluffy-yarn.jpg" alt="On Saturday, Fluffy was playing with a
> ball of red yarn in my living room.  It was quite funny.">
> <!-- no following paragraph -->
>
> In this context, @alt is necessary to provide equivalent content for
> the image.  Completely replacing the image with the @alt content
> doesn't change the meaning of the document.

Agreed here too. However, I think the first example might be better  
practice since it includes important information like "Saturday" and  
how the author found it "quite funny" that should be available to all  
users and that probably isn't apparent from the image itself.

> <object data="abbott-costello.mpg" title="Who's on first?">
> <object data="abbott-costello.ogg"></object>
> </object>
>
> Here, the outer <object> has fallback to another <object>.  That is
> another case of equivalent content - providing equivalent media in
> another format.  But what should the contents of the inner <object>
> be?
>
> I contend that it should be the full transcript of the "Who's on
> first" sketch.  That would also be equivalent content.  I honestly
> believe this will be rare - authors will rarely provide a truly
> /equivalent/ textual alternative for a video.

I don't think this is a good way to use equivalent (or at least its  
not compatible with how WAI and other in the community are using it).  
When @alt is used, authors are expect to include an equivalent  
description if it can be kept brief. If it can't than @alt should  
still be used, but @longdesc should be used as well. I think there  
are problems with using the term 'equivalent", but I can also see why  
it's encouraged.

 From the author's point of view, every nested <object> after the  
first is a fallback: i.e., it is less preferred by the author. If it  
was merely a matter of providing alternatives to the user and the  
user agent, this would be better accomplished through content  
negotiation. Once you arrive at the final <object> element's content  
this is an equivalent (or alternate) in the same sense as the other  
nested <object> elements. The author would prefer to vend the first  
object, however, the author has provided these ordered fallback  
equivalents (equivalent in the sense that the author has strived to  
provide content that still conveys the same meaning as the author's  
first choice; all the nuance of the first choice may be lost so  
there's no equivalents in that sense). The classic example I think is  
a group of nested objects with:

1) a java applet interactive game where you can be either Abbot or  
Costello
2) a video of the sketch with Abbot & Costello
3) a still image of some moment in the sketch
4) a textual description of the sketch (or a complete dialog as you  
suggested).

In some sense these may all be equivalents (and equivalent in the  
sense used by WAI). They are not equivalent in the sense that the  
author would like you to see one of the alternates and prefers the  
order listed (the order of the nested <object> elements).

Just to bring up @alt again, the issue is how to handle <object>  
fallback when it is very lengthy? Is there still a need for a brief  
description to serve as a placeholder on the page (one that is  
readable and also maintains the CSS layout for example). Or should a  
UA simply extract the first n words of the long description. Or for  
aural UAs would it be better to simply allow the user to begin  
hearing the full contents (rather than previewing the @alt attribute)  
and then let the user cut-off the UA at any time and move on. Again I  
don't have answers to these questions, I just think they're important  
questions to pose.

> Otherwise, it should be a short (possibly rich) description of the
> sketch, with possibly a link to more information.  But a description
> of the content is not the same as an equivalent to the content, so the
> UA should let the user know that the media is missing, why, and
> possibly provide a link to the missing media. (The user could be
> sighted, but not see the media for other reasons.)

The contents of the last <object> element could also be a lengthy  
description of the sketch just not the full dialog). Would that not  
also be an equivalent? Or for an audience that the author expect to  
all know what the sketch was about the phrase: "Abbot & Costello's  
'Who's on First' sketch" could serve as an equivalent even though it  
is a breif description. On the other hand, even the entire dialog is  
not necessarily an equivalent if the sketch involved visual comedy or  
improvisational parts not included in the scripted dialog. In some  
ways I think we end up way too deep in these subtleties if we try to  
differentiate between equivalent text and non-equivalent (but still  
alternate) text. My understanding is that the alternate should be  
equivalent (in the way WAI uses equivalent). I think it is better to  
make the distinction between abridged and unabridged alternate (and  
also equivalent) text, but not whether its equivalent versus  
descriptive.

Also, as I said above, even the 2nd nested <object> is not an  
equivalent in the sense the author viewed it as a fallback. It is not  
the author's first choice media format. It is not only when you reach  
the text media that it ceases to be an equivalent in that sense. In  
this sense it might make sense for the UA to let the user know  
whenever they're not consuming the authors preferred content. In the  
example I gave above, the still image may even be a worse fallback (a  
lesser equivalent if that's not too Orwellian) than the full text of  
the dialog or even a lengthy description of the dialog.

> These two possible textual "alternatives" should have different markup
> for a couple reasons:
> - So authors don't confuse "equivalent" content with "descriptive"
> content the way they do now (I would still like to know if there's a
> decent way I can gather statistics on how "alternate content" really
> is used on the web.)
> - So UAs can present "equivalent" content as a drop-in replacement for
> the media, but "descriptive" content as an accessible fallback with a
> note that certain media are missing.

On this second bullet point (and despite my misgivings about the  
terminology) isn't this what adding @alt to all embedded content  
elements would accomplish? In your words it provides two different  
markups so that "UAs can present 'equivalent' content as a drop-in  
replacement to the media [the contents of <object>], but  
'descriptive' content [as the value of an @alt attribute] as an  
accessible fallback with a note that certain media are missing"

>
> Hope that better says what I meant.
>
> I really think <img> is useless for images that are "worth a thousand
> words", so I'd prefer to use <object> better than try to fix img/@alt
> and img/@longdesc

I understand. I think that is the solid use-case/problem that we're  
all trying to solve. It may be we end up endorsing many solutions.

Take care,
Rob

Received on Monday, 16 July 2007 00:43:56 UTC