W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > October to December 2005

Re: Validation as test for basic accessibility

From: Maurizio Boscarol <maurizio@usabile.it>
Date: Thu, 10 Nov 2005 12:31:45 +0100
Message-ID: <43732FA1.5000900@usabile.it>
To: "Bailey, Bruce" <Bruce.Bailey@ed.gov>
CC: w3c-wai-gl@w3.org

Bailey, Bruce wrote:

>>>22.  The evidence is that the one-way statistical correlation 
>>>between sites that are valid to those that are accessible is 
>>>too overwhelming to be explained away by the hypotheses offered.
>>This is totally undemonstrated as far as we know.
>You are mistaken.  The actual hypothesis is something like:  "Paying attention to *only* validity results in highly accessible web sites."  

Who decided this is the actual hypotesis? It's not mine.
I've already stated my hypotesis:
"Validilty is a preliminary, necessary requirement to accessibility"
We need to try to falsificate it. If we can't, we can hold the 
hypotesis. If we can falsify the hypotesis, then we can't accept it.

This is how research is supposed to work, at least.

Unfortunately, we already agree that the hypotesis is not always true, 
so we can't hold it.
This is the hypotesis we discuss, this is what we have to decide whether 
to put in WCAG or not.. I think we're talking about different things.

>>The correlation may be caused if the careful, professional 
>>developers who choose to attend to validity are more likely to 
>>also attend to accessibility.
>That is a rational hypothesis but, in light of the observed experiential data, it must be regarded as an extraordinary claim requiring *some* evidence.

We're definitively talking about different things. I could state that 
developers who attend to validity are more likely to attend to 
accessibility. Sure they are! This is because the places (forums, 
webistes, mailing lists) where we talk about accessibility we often also 
talk about validation and vice versa. And because wcag 1.0 already made 
the correlation between acc. and val.
And because if one has the attitude to inform himself about rules, he is 
more likely to discover even more rules... This is so obvious that I 
can't talk about this a second more.

But, again, it's not what I'm talking about.

The point of my discussion is not if validation is or is not a best 
practice for web designer: it is. I strongly reccomend it.
As a process, validation is extremely useful.

But as a product quality, is validation necessary to have an accessible 
website? (see hyp. above)

If we aren't able to distinguish "product standard" and "process 
standard", we're not walking a long way.

As an example, ISO 13407 is a process standard: it explain how to 
develop things (using user centred design, in that case).
ISO 9241 (and even ISO/IEC 9126) are product standard: they define the 
requirements products must have to reach a stated quality level. They 
talk about interface features, code quality, ando so on.

In wcag maybe we are sometimes mixing up product and process. As a 
process (as I've already said in many other post), there are some 
practice that increase the probability to have a quality product.
But Wcag will also be used as a testing methodology on product features. 
It will measure if a product is accessible, not (not only) how it has 
been made.
Now the problem is: validaton is great automatic tool to develop and I 
use it as often as I can. But is any invalid site, at the end, really 

Well, we already agree that it's not always true. So, if it's not always 
true, it's not true as an absolute statement. We can't hold that statement.
But, if we put validity at L1, our future test based on wcag 2.0 will 
conclude that every, even trivially, invalid pages and sites would be 
not accessible, even if they wouldn't be.

This is the error we are now deciding to authorize. It's up to us, of 

The question then may be: how high is the risk of excluding accessible 
site from our L1 level test to be declared L1-compliant even for trivial 
(irrelevant for actual accessibility) errors of validity? If we think 
this risk is very low, we can forget it. But if the actual risk is a 
little higher than we think, we're making a low-power test from our set 
of guidelines. By "power of the test" I talk on the analogy of 1-beta 
(probability of 2nd type error) in statistical analysis. Just for the 
ones that know about it.

We shouldn't forget that our WCAG will be used for a variety of 
different goals. Not only defining a quality process, but also to set up 
test tools, and a set of laws. Things changes, when you define a quality 
process and when you write mandatory legal requirement, or when measure 
the level of something.

The problem that generate this discussion may be the mixing up product 
and process features, I suspect.

Received on Thursday, 10 November 2005 11:19:02 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 16 January 2018 15:33:57 UTC