Re: ACTION-128: Draft @summary voting text in conjunction with PF

On Jul 6, 2009, at 1:16 PM, Ian Hickson wrote:

> On Mon, 6 Jul 2009, Maciej Stachowiak wrote:
>>
>> When you guys develop the wording, I'd suggest also having this as an
>> explicit option:
>>
>> ( ) The text currently in the HTML5 spec is mostly fine, but should  
>> in
>> addition make summary="" conforming for authors without encouraging  
>> it over
>> other options:
>>       http://www.whatwg.org/specs/web-apps/current-work/multipage/tabular-data.html#table-descriptions
>>
>> I believe this represents the point of view of many Working Group
>> members better than the first two options, enough that it shouldn't  
>> be
>> just a write-in.
>
> I'm fine with having that as an option if we do have a vote, but  
> wouldn't
> allowing the use of this attribute fail to resolve the main issue  
> being
> raised here? Namely that authors don't understand what the text  
> should be,
> and if they use summary="", they don't see the text and don't fix it.

First of all, I agree that preponderance of the evidence shows  
summary="" to be an ineffective feature, as currently deployed.  
Despite some casting doubts on the methodology of studies done on  
this, it seems to me that all the data points in the same direction,  
and there hasn't been any material counter-evidence presented, only  
armchair reasoning. Many of the criticisms also strike me as off-base  
- for example, it's not usually considered a scientific argument to  
call a statistical sample invalid because the person gathering the  
data is believed to be "biased" (i.e., is predisposed to believe a  
particular hypothesis). Rather, you have to point out actual flaws in  
the sampling method. So at least to me, it's pretty well established  
that summary="" is not a great solution for its use case.

What I don't think has been established to the level of scientifically  
persuasive evidence is that making the attribute nonconforming for  
authors is the only way or the best way to improve the situation in  
the future.

I believe the issue would be resolved by recommending alternatives  
that are likely to be superior, and through emitting optional warnings  
in validators (for example, when validator.nu sees a summary=""  
attribute it could recommend that you review whether the text is  
better presented to all users, or if perhaps it does not add any value  
for users of screen readers). Or at least, that seems like it would be  
about as effective as making the attribute nonconforming for authors.  
The advantage would be that advanced experts could still make a  
considered judgment that the attribute is the right choice in some  
cases, even if most authors are guided towards explanatory text that  
is visible to everyone.

> Why would we continue to allow people to use an attribute that we know
> they get wrong all the time?

I think highlighting effective alternatives plus optional validator  
warnings will reduce the incidence of incorrect use. If I'm wrong and  
these measures are not effective, then I think making the attribute  
nonconforming would be similarly ineffective.

Or at least, that is my position. I think it is reasonable to disagree.

Regards,
Maciej


Regards,
Maciej