Re: Images and alternative text

Ian Hickson wrote:
> On Thu, 7 Aug 2008, Philip Taylor wrote:
>>   <img src="..." alt="{x \over y} = {1 \over {y \over x}}">
> 
> ...would be a horrific alternative text to give a screen reader.

I'm not aware of any better alternatives for the general case. Text like:

>    <img src="..." alt="The fraction x over y is equal to 1 divided by the 
>    fraction y over x.">

is clearly unscalable when you have non-trivial equations.

Since these LaTeX-to-HTML tools are primarily aimed at sighted users 
with graphical browsers, and only a tiny fraction of users will rely on 
screen readers, it doesn't make sense to spend significant effort 
optimising the output for screen readers. The tools can't be expected to 
ask authors for a textual alternative to each equation (because the 
whole reason for this new alt stuff is that tools often can't do that 
and have to make do with the limited information and intelligence they 
have). So the tools will carry on doing the best they can do that 
doesn't take much effort, which is to put the LaTeX source of the 
equation in the alt attribute. Given that that situation seems unlikely 
to change, it's a problem if HTML 5 adds features that will trip people 
up by reserving certain syntax and making it non-conforming to match 
that syntax when supplying real alternative text.

(If it wasn't non-conforming, then there would be less of a problem - 
people could write <img alt="{x \over y}"> with no worries other than 
that it might be rendered slightly differently than other equations by a 
few UAs, which isn't a major issue; or they could write <img alt="{LaTeX 
equation: {x \over y}}"> so it'd at least be handled consistently for 
all equations. But HTML 5 doesn't allow either of those, so the 
conformance requirements are a problem for people who care about 
conformance. But making it conforming still wouldn't solve the problem 
of some UAs misidentifying the equations as critical images with no 
attempt at proper alternative text.)

> I don't think it's equivalent to the image at all. It's the source of a 
> program that was used to generate the image, but that's not the same 
> thing. Would you consider the replacement text of a fractal to be the C 
> source code that generated it? Or the replacement text for an SVG file to 
> be the raw source code of that SVG file?

The mapping between source and output in those cases is a very 
substantial change - the text "<circle>" and the graphical rendering of 
a circle are very different, and information that is obvious in one 
format can be almost entirely hidden in the other. The mapping between 
LaTeX source and output is far more straightforward - the text "{x^2 
\over y}" and the graphical rendering

    2
   x
  ---
   y

are quite similar, and it doesn't get much worse when the equations 
become more complex. The changes are just in layout, and in removal of 
some symbols, and in rendering of some symbols as special shapes - 
there's no revelation of hidden information in one form or the other[*]. 
So I don't think the analogy works.


[*] (except in some quite rare cases of mathematical jokes, like 
"\lim_{\omega \rightarrow \infty} 3 = 8", which only make sense in the 
graphical form)

-- 
Philip Taylor
pjt47@cam.ac.uk

Received on Friday, 8 August 2008 13:44:33 UTC