Re: ISSUE-30 (Longdesc) Change Proposal

Jonas Sicking On 09-10-29 22.47:

> On Thu, Oct 29, 2009 at 2:15 AM, Leif Halvard Silli:
>>> And regarding descriptions:


>> I take it that in your view neither aria-label, aria-labelledby nor
>> aria-describedby represent a [near] duplicate feature of @alt.

    [...]

> But if it matters what I think regarding it's [near] duplicity: I do
> think that @alt seems intended to be used different from
> @aria-label/@aria-labelledby/@aria-describedby.


Agree.

> However in practice I
> think you'll find a significant amount of @alt attributes actually
> containing information that is a description rather than fallback.


Probably.

But there are at least 3 definitions of "fallback" in the air here 
...

I am in favor of the most literal: replacement content that is 
kept inside the element at hand. ALT falls in this category. And, 
longdesc (IMHO), even if the actually long fallback text is only 
represented as a link. (The typical example of longdesc content 
could be diagram image presented as a table.)

You understanding, as presented here, is touching the quality or 
essence of fallback content: If isn't of a certain kind or 
quatlity, then it isn't fallback.

ARIA-describedby etc represents a third kind of fallback: the user 
agent computes a text equivalent based on things that the author 
has linked to the object - directly and/or indirectly. ARIA uses 
the wording 'textual equivalent' - which is a quality/essentiality 
understanding of the result of that computation.

But anyway, I think that the reason that the WAI consensus 
document accepts aria-labelledby as a replacement for @alt, is 
precisely because - often - the quality will be the same as for 
real fallback - @alt. (You could say that it represents an 
acceptance of the fact that not all @alt content is tip top.) 
That's how I evaluate it.

> In
> fact, I suspect there is more such usage of @alt, than fallback usage
> of @alt. But I don't have data either way.
> 
>>> The same can not be said for @longdesc and @summary, neither of
>>> which has seen any significant amount of real-world uptake.
>>> Yes, there is more than zero uptake, but I don't think there is
>>> enough to warrant having duplicate (or near-duplicate)
>>> features.
>>
>> Above you lead us to see that the neither aria-label, aria-labelledby nor
>> aria-describedby are fallback, and thus are not [near] duplicate features of
>> alt.
>>
>> In my previous reply, I gave you the spec definition of @longdesc that you
>> asked for: It is a long variant of alt. Thus it is fallback.
>>
>> Therefore I am baffled that you bring up "duplicate feature" again.
> 
> Ah, I think I missed to reply to this argument before.
> 
> What do you have to back up the claim that @longdesc is fallback?
> 
> I did some research for this statement and here is what I found.
> 
> Support from specs:
> 
> The DTD from the HTML4 spec says for @alt
> 
> "short description"
> 
> and for @longdesc
> 
> "link to long description (complements alt)"

Do we need to discuss if @alt represents fallback? If @alt is 
fallback, and if the longdesc resource complements @alt then it 
seems to me that it is meant to represent fallback.

> So here it seems to indicate that both are descriptions, and that
> @longdesc is a longer version of @alt. Note that I don't think that
> you can infer from "complements alt" part that the two have similar
> functionality.

I note it. But I don't understand it.

Let's not get caught in words. You said that @alt is described as 
description. Should we infer from that that it is not fallback? 
You can just turn of image support in your web browser and verity 
that it falls back to display the textual content.

Of course, @longdesc isn't fallback in that sense - as the UA 
doesn't fall back to the long description. But conceptually it is 
a manual, user activated fallback. It is meant to lead to a 
resource that more fully represents the image of the IMG element 
in an accessible way. And the fallback link /is/ kept inside the 
IMG element itself. So if you as user are unable to make sense of 
the image, you have option to fall back to the long textual 
description/version/fallback/representation of/for the image.

> However, if you look at the actual descriptions of the two attributes,
> it says that about @longdesc
> 
> "This attribute specifies a link to a long description of the image"
> 
> So far this matches the DTD. However for @alt it says
> 
> "For user agents that cannot display images, forms, or applets, this
> attribute specifies alternate text".
> 
> I.e. here it seems to talk about @alt as fallback.

Yes, and in the technical sense that I prefer.

> So it seems to me that the HTML4 spec is internally inconsistent, or
> uses different terminology from ARIA. I.e. possibly it treats
> "description" and "alternative text" as the same thing.

I think so too.

> So there's two ways where I can see that you could argue that HTML4
> defines @longdesc as fallback:
> 
> 1. If you read the part where it says that @alt is fallback. And also
> read the part where it says that @longdesc is a longer description and
> @alt is a short description, and look at the fact that that would make
> @longdesc a longer version of @alt, but ignore the fact that it says
> that both are descriptions.


Yes. Though I don't ignore "description" completely, because I 
don't support the rigorous interpretation of 'fallback' as 
'textual replacement'.

> 2. If you argue that everywhere where HTML4 says "description" it
> means "alternative text" and thus "fallback".


HTML4 probably isn't that consequent. ;-)

> Support from implementations:
> 
> I can't find any implementations that treat @longdesc as fallback. No
> browser that I know of displays the contents of the uri that @longdesc
> points to when you turn off images. None even displayed a link or the
> @longdesc uri.

I agree that they don't technically fall back to longdesc.

However, the fallback link is kept inside the IMG element. Thus it 
is represented inside the element. You can link to the IMG, and 
simultaneously you also include the link to the fallback. Also, we 
are talking about a two level fallback: You get the alt first. But 
can opt for the entire cake if you wish.

I have also explained what I meant by saying that UAs treat it as 
fallback: They make the @longdesc link as much of a fallback as is 
possible. The fact that the link opens as a link with 
target="_blank" is what does the trick, so that you don't loose 
focus of the main text just because you read the linked resource.

> However I only checked a few popular browsers. Does anyone know any
> implementations that treat @longdesc as fallback


If you accept my description of how longdesc is fallback from 
above, then I can mention Opera, iCab and Internet Explorer in 
combination with JAWS. Plus, according to Charles's reply, the 
Firefox extension.

We can put it like this: If a user agent supports activating the 
longdesc URL, then it supports longdesc in a fallback way. No more 
and no less. Where is it said that the fallback can't be a link? 
Doesn't HTML 5 recommend that <video> could contain links to where 
you can download the proper codec etc?

@longdesc is also part of the DOM specification and can thus be 
activated in literally all JavaScriptable browsers. We don't  need 
to consider JavaScript support as support, but we shouldn't forget 
it either.

> Support from the actual change proposal from Charles:
> 
> It appears clear to me that Charles is proposing @longdesc to be a
> description, not fallback. Quoting from his change proposal:
> 
> 4.8.2.1.1, 4.8.2.1.2, 4.8.2.1.3 should all mention that a longdesc
> *may* be provided to provide a detailed *description* of the image,
> e.g. to help a person who cannot see it to find it from a description.
> 
> (emphasis is part of the actual change proposal).


I'm sure Charles will explain. But some pictures are indeed that 
complicated that a description is what is needed, if the fallback 
is supposed to be any useful. And I cannot imagine that Charles 
excludes e.g. tables as substitute for image diagrams etc just 
because he used the word 'description'. And what is a description 
- is there a definition of that? Let's not get lost in words.

> All in all I could not find a good case for @longdesc being fallback
> rather than a description. The best case was the HTML4 spec, where you
> had to either use extremely selective reading, or had to assume that
> it was confusing terms.


That HTML4 could have been more precise is certainly correct. But 
I disagree that that one must wear any special kind of hat in 
order to make sense of what HTML 4 says about @longdesc.

> However, ultimately, if you think that @longdesc should be used to
> indicate fallback I suggest that you make a Change Proposal to that
> effect. It does not appear that this is what Charles is proposing.

Charles, do we disagree about anything here, you think? I don't 
think so, really, this disagreement appear to me to be a bit 
constructed.
-- 
leif halvard silli

Received on Thursday, 29 October 2009 23:14:03 UTC