[uf-discuss] Re: Precise Expansion Patterns

[Ben Ward]

On 16 Dec 2007, at 20:09, Manu Sporny wrote:
> It is important for us to focus on the reason this discussion  
> started in
> the first place:
>
> http://microformats.org/discuss/mail/microformats-discuss/2007- 
> December/011035.html
>
> The issue was accessibility, specifically, how accessible is the ABBR
> design pattern for those that use screen readers.

No, Manu, that was not the reason this most recent discussion started.

In fact, the catalyst for this most recent iteration concerns not  
accessibility — I deliberately avoided that as finding precise data  
is too difficult. The issue at hand is that more recent  
specifications such as GEO (albeit brainstorming) and hAudio are  
mandating the use of the ABBR pattern in a way which is not  
compatible with the HTML specification.

Yes, there are many here who care a great deal about the implications  
of microformats on users of assistive technology, but it is clear  
that most contributions here are unable to find sources or recorded  
evidence to support or refute any claim. Unfortunately, gaining such  
evidence from people who really use AT daily is neither easy nor  
inexpensive. You or I downloading a trial of JAWS and running it will  
not useful test results.

> This is not true. You can set several, of not all, screen readers  
> to not
> read titles of SPAN elements.

The issue is not whether you _can_ set a screen reader to read or  
ignore @title attributes, it is whether users actually do or not. The  
limited experience I have from inside Yahoo!, where I have been able  
to ask some very generous people to assist in accessibility testing  
on another issue, is that people who depend upon AT tools are far  
more inclined to customise their tool to improve their experience. As  
such, there are a plethora of combinations of tools and  
configurations consuming pages.

One can presume on the basis that these users are more inclined to  
configure their tool, that such a user will configure their tool  
optimally for their usage, depending on the kind of content they  
interact with the most. As such, we cannot ever work on the basis  
that upon discovering machine data in the @title attribute of a  
microformat property that they will simply reconfigure their tool;  
their choice to enable reading of titles will be useful for some  
kinds of content.

It is the quantity of variables in the field of AT and the expense of  
testing them which makes it hard for a community of our limited  
resources to make decisions based on AT performance. But whether  
criticising or supporting a pattern, vague statements about the  
behaviour of AT help nobody.

I think this discussion would progress better if people stay focused  
on the data requirement and the semantics of the output first, and  
the implementation second. So far, we're getting very sidetracked by  
a series of new proposed hacks, rather than identification of which  
problems need solving by a precision/expansion pattern.

Thanks,

Ben