W3C home > Mailing lists > Public > www-validator@w3.org > September 2013

Re: Error on generator-unable-to-provide-required-alt

From: Jukka K. Korpela <jkorpela@cs.tut.fi>
Date: Tue, 03 Sep 2013 12:43:01 +0300
Message-ID: <5225AF25.1040204@cs.tut.fi>
To: "Jan P." <proedie@arcor.de>
CC: www-validator@w3.org
2013-09-03 12:20, Jan P. wrote:

> Am 02.09.2013 13:42, schrieb Leif Halvard Silli:
[...]
>> Thus, if the validator *is* reporting lack of alt as an error, then
>> there is no permission to accept the
>> @generator-unable-to-provide-required-alt attribute. And if the
>> @generator-unable-to-provide-required-alt attribute is present,
>> validators ar still permitted to report lack of @alt as an error.
>
> And this does not make any sense at al.

I strongly agree, but the culprit is the nonsensical attribute 
generator-unable-to-provide-required-alt rather than anything else. But 
this is a matter of HTML5 specifications, not validation.

> So what could I do? Writing nonsense into the alt value to satisfy the
> validator?

Why would you need to satisfy the validator? The validator is there to 
serve you, not vice versa.

What matters is how you serve your potential visitors. For this reason, 
you should include useful alt attributes when feasible. This means 
writing the obvious text when an image represents text or a symbol that 
has an easy text equivalent. The rest is rather controversial. But in 
general, any alt attribute that gives at least a cue about the content 
of the image (such as alt="(Photo of a silly walker.)") is usually 
better than no alt attribute, which in turn is better than an absurd or 
patently wrong alt attribute.

So the validator is useful in detecting img tags that accidentally lack 
alt attributes. It is not useful in a debate over the general necessity 
of alt attributes.

> In a nutshell, this is a *feature request*, not a bug report: When there
> is a generator-unable-to-provide-required-alt="" attribute, the
> validator should report a warning instead of two errors.

While it might make sense to issue one message only, I don’t think any 
time should be sent in improving a feature that should not be there at 
all. Besides, not reporting an error when there is an easily detectable 
violation of the specification (the nonsensical attribute) does not 
sound like validation to me.

Yucca
Received on Tuesday, 3 September 2013 09:43:30 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 3 September 2013 09:43:37 UTC