img/alt summary attempt

Is there agreement on at least the following?

(1)  Web pages SHOULD be accessible to all.

(2)  Much of the improvement in alt over the past years is due to
social pressure.

    (2a)  HTML standards can be a force better accessibility.

    (2b)  But standards are a very weak force -- requiring anything
beyond interoperability will reduce their effectiveness.  [I know
there isn't agreement on whether alt is worth the price -- I'm just
confirming agreement that such a price exists.]

    (2b)  This social pressure has included legal requirements -- it
is easier to defend the time spent writing alt attributes if there is
a law requiring accessibility and a standard stating that alt is
required for accessibility.

(3)  WCAG is the source of normative accessibility information from the W3C.

    (3a)  The HTML specification should at least reference WCAG2, and
warn that a missing alt, even if adequate for HTML5, is not adequate
for WCAG2.

    (3b)  There is disagreement on whether conformance checkers should
continue to flag missing alt attributes.

(4)  alt SHOULD represent an equivalent to the picture, when possible.

-- Note -- HTML5 expands the use of LEGEND -- there may be cases where
alt is not useful *because* it is redundant with LEGEND.  This case
may not be covered in current WCAG Recommendations.

(5)  There is disagreement on what to do when alt cannot be a true equivalent.

(5a)  WCAG currently requires a description of intent, because that is
better than nothing.

(5b)  The current HTML5 drafts recommends leaving alt out completely,
because machine-generated alt texts may be worse than nothing.

(5c)  Common current practice is to either leave the alt out (and not
worry about validity) or to set alt="" (and claim that the image is

(5d)  Several people have suggested specific magic values to indicate
_decorative or _notsupplied.  This is more explicit, which should at
least separate the reasons for missing alt.  Except that the people
who care enough to do this may not be the ones writing problem pages,
so it still might not work...


Received on Monday, 14 April 2008 17:48:51 UTC