Re: CP: role=img trumps empty alt (ISSUE-206: meta-generator)

On Sat, Aug 4, 2012 at 5:11 AM, Leif Halvard Silli
<xn--mlform-iua@xn--mlform-iua.no> wrote:
>> I don't think we need "a signal that the image is significant but in
>> lack of alternative text" as:
>>
>> 1. Developers can already distinguish images that are not significant
>> by adding @alt and images that are significant by omitting @alt.
>
> Yeah, and that is a quite complicated thing: Add something if it is
> _not_ significant, and remove something if it _is_ significant.

Not "remove" something, just don't robotically add it.

> Especially so, when the starting point is <img> elements that have an
> empty @alt in order to validate.
> The problem we try to solve is this: Help generators signal that
> something is "significant" without being stamped as invalid.

Or, at least, _might_ be significant. I think proposals that allow
developers to not robotically add @alt are much stronger here, such as
@relaxed, not least because they benefit from existing interpretation
by user agents.

> The problem with an <img> that lacks the @alt attribute is that this
> may or may not be an error. (It is only in very specific cases that it
> isn't.) Thus, omitting is not a general solution. If it were, then we
> would not need @incomplete/@relaxed either.

@relaxed is a signal to linters; it's not a signal to user agents.

>> 2. AT interested in exposing images that might be significant to end
>> users already need to expose images with missing @alt.
>
> Since this proposal suggests that generator developers should continue
> to use empty alt, then it necessary to add role=img if we want to make
> them interested in exposing it.

That your proposal suggests adding alt="" is a defect of your
proposal, as it has a much worse backwards compatibility story.

--
Benjamin Hawkes-Lewis

Received on Saturday, 4 August 2012 09:50:32 UTC