Re: Images and alternative text

Boris Zbarsky 2008-08-11 05.34:

> Leif Halvard Silli wrote:
>>> Thing is, in some communities sending TeX as part of your text 
>>> message is considered perfectly reasonable, at least for simple 
>>> equations.  So a lot of people don't think of it as "source code" but 
>>> as simply another notation, one well-suited for plaintext communication.
>>
>> Yes, apparently this is so. But I consider that to be a "private 
>> agreement" and not part of HTML.
> 
> So is anything that uses any words that aren't common knowledge. 
> Whatever "common knowledge" means.


Of course. However, HTML 4 and 5 and WCAG 1.0 and 2.0 have rules 
for how to use IMG, and for the relationship between the image and 
the @alt text.

It strikes me that the math equivation use case has not been 
mentioned at all in the WCAG 2.0 guidelines (as far as I could 
spot). Yet it is and important example of when, what WCAG 2.0 
calles "images of text", is necessary.

WCAG 2.0 says that images of text should only be used if a 
"particular presentation of text is essential to the information 
being conveyed" [1].  And I suppose that WCAG would agree that the 
text images are essential in the case of math equivations. I also 
assume that WCAG would recommend using TeX rather than the images, 
if the TeX code really was a full replacement ...

Btw, saying that the code is TeX would not help me understand it 
(though it could help the UA). It only meant that I could find a 
method to understand it. Wheras uncommon words does not have this 
problem. I know that it is a word. And I know several ways to 
learn words. Wheras a special syntax "out of the blue", would 
require Google or some such tool to learn what it is. ;-)

 
>> I am against the fact that there is nothing in the source code of e.g. 
>> Wikipedia or MathML 2.0 (which follows the same cowpath - plus it also 
>> uses <blockquote> for indenting) which informs us that this is LaTeX.
> 
> I am by no means opposed to extending <img> to be able to indicate more 
> metainformation about its alt text.

Good. I wonder what the best method would be. Using lang="zxx" 
would merely offer a hint. Perhaps we would need to look at @role 
again?

>> There has been some debates here about whether User Agents, in 
>> particular screen readers, should rely on heuristics to detect what 
>> should have been written in @alt etc.
> 
> When the @alt is missing, right?

Yes.

>> But here in _this_ case, suddenly we hear from what I am so free as to 
>> characterize as "the other side", that putting LaTeX source code into 
>> @alt - just like that - is OK.
> 
> I sense that you're implying there's a contradiction here, but I'm sort 
> of failing to spot it, if so.

The similarities are these: When @alt is lacking, one must apply 
the heuristics directly on the image. When the @alt contains TeX, 
one must guess what the source code contains (or remain baffled). 
It would be logical to say that in neither case should one be left 
to guess what the image - or the @alt - contains. To reject/accept 
noALt, while at the same time *not* also reject/accept the use of 
TeX, without telling that it is TeX, appears to mea as being 
inconsequent.

>> There should be a microformat or something, for how to put LaTeX into 
>> @alt.
> 
> That's a distinct possibility (though note that the core issue is not 
> limited to LaTeX).


Right. And as this thread really is about when and how to fill 
@alt with keywords instead of offering a full @alt text 
equivalent, we also need a "microformat" for signaling the use of 
such keywords. At least, such is the presumtion which has lead to 
the curly brackets syntax, I think. I proposed @tagged, and later 
role="album-photo tagged", for this. And it appears as if e.g. 
@role could be used for a "microformat" in both cases.

However, perhaps the 'tagged' keyword should contain 'alt', to 
signify that it is about @alt: role="album-photo alt-tagged". 
Then, for the usecase of "images as text", I could - in the case 
of math equations - imagine <img role="textimage 
alt-tex-sourcecode" ... > . For instance.

[1] See point 1.4.5 under this link: 
http://www.w3.org/TR/WCAG20/#visual-audio-contrast
-- 
leif halvard silli

Received on Tuesday, 12 August 2008 01:00:00 UTC