W3C home > Mailing lists > Public > public-html@w3.org > August 2008

Re: alt attribute - a very simple proposal

From: Henri Sivonen <hsivonen@iki.fi>
Date: Wed, 20 Aug 2008 09:23:53 +0300
Cc: Ian Hickson <ian@hixie.ch>, HTML WG <public-html@w3.org>
Message-Id: <ADDB3106-1F8A-41B3-97CE-538186C7849A@iki.fi>
To: Karl Dubost <karl@w3.org>

On Aug 20, 2008, at 05:13, Karl Dubost wrote:

> Le 19 août 2008 à 22:03, Henri Sivonen a écrit :
>> On Aug 19, 2008, at 15:24, Karl Dubost wrote:
>>> My proposal is something along:
>>> 	All img elements must have the alt content attribute set.
>>> 	The accessibility requirements on the possible values of
>>> 	the alt attributes are defined by WCAG 2.0 and not HTML 5.
>> When the markup generator has an image of unknown (to the  
>> generator) content for which the generator does not have user- 
>> entered textual alternative, what should the markup generator emit  
>> as the value of the alt attribute according to your reading of WCAG  
>> 2.0?
> Maybe two possible scenarios:

Could you point me to chapter and verse where you found the answers in  
WCAG 2.0?

As far as I can tell, WCAG 2.0 doesn't address this case at all. WCAG  
2.0 covers how to make content accessible when the person acting on  
WCAG 2.0 advice has access to all content on a site and is willing to  
spend non-trivial time making things accessible.

It seems to me that WCAG 2.0 doesn't address the issue of what the  
HTML generator writers of media sharing sites should do to minimize  
inaccessibility when the users of the media sharing site *won't*  
participate in making their uploads accessible.

And that's one the problems with just referring people to WCAG 2.0:  
WCAG 2.0 only covers an accessibility-oriented subset of Web  
authoring. HTML needs also address the inaccessible part of the  
spectrum of HTML usage in a realistic way.

(I'm stating some of my fundamental assumptions for clarity. I think  
these should be intuitively reasonable, but obviously people on the  
list disagree.
  1) HTML needs to make authoring accessible content possible. (This  
is different from requiring all HTML content to be accessible.)
  2) Media sharing sites are a notable part of the Web and, therefore,  
the HTML spec should support media sharing sites.
  3) Supporting a use case means providing conforming aka. valid  
syntax for the use case.
  4) The validity definition should be formulated so that editing a  
page from not valid to valid through the path of least resistance  
doesn't make the accessibility properties of the page worse.
  5) The validity of HTML generated by a media sharing site should be  
under the control of the programmers of the media sharing site. User  
input must not render the HTML output of a media sharing site invalid.
  6) Private persons uploading content to media sharing site will not  
in many cases provide textual alternatives for the media they upload.  
Arguments about whether they *could* are moot if they in reality  

I think it follows that to satisfy the above, valid HTML emitted by  
media sharing sites will not always be accessible. The programmers of  
the media sharing site can modify the result only to make the output  
not valid in some cases. It's not in their power to modify the  
situation to make the output accessible in all cases.)

The other problem with just referring people to WCAG 2.0 compared to  
the HTML5 spec explaining alt usage itself is that WCAG 2.0 covers  
fewer authoring situations for alt than the HTML5 draft and contains  
no examples whereas the HTML5 draft has examples. Moreover, instead of  
being a self-contained document, WCAG 2.0 itself is very abstract and  
comes with an separate exegesis document (Understanding) which still  
does not contain examples. To find examples, one needs to go find a  
*third* document (Techniques) and search it for relevant techniques.

I understand that just referring to WCAG 2.0 would be a politically  
simple solution, but would it really make HTML5 a better spec?

> 1. The generator creates the markup with an empty alt attribute.
>   Rationale: any kind of automatic machine value attribution will be  
> unsastifying.
>   1.a (Optional) the markup generator generates a report about  
> possible image issues.

This fails #4 above.

> 2. The generator creates no markup for alt attribute and leave it to  
> the user to fix it later on.

This fails #5 above if we accept the part of your proposal that makes  
alt syntactically required.

Henri Sivonen
Received on Wednesday, 20 August 2008 06:24:36 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 29 October 2015 10:15:37 UTC