Re: Investigating the proposed alt attribute recommendations in HTML 5

At 23:46 -0500 UTC, on 2007-09-03, Robert Burns wrote:

[... a lot of text snipped]

Robert, I cannot follow most of this message. In part because you quoted so
much that I have no idea what exactly you're responding to. If you'd quote
only just enough to provide necessary context (or provide short summaries of
snipped text, as you can see me do in most of my messages), it gets much
easier for people to understand what you're saying.

I'll try to respond to the bits I think I can follow, and indicate what I
can't follow.

> I think saying that authors need to think about what they want to
> convey with a piece of embedded content.

Something seems to be missing here.

> However, I do not think it
> is helpful at all to say that it shouldn't be a description of the
> image. If the embedded content has anything to do with what the
> author wants to convey in including the embeeded content in the
> document - i.e., its not just purely decorative in a completely
> unrelated way - then the meaning the author wants to convey with the
> embedded content will be a description of the embedded content (in
> some sense).

Eh?

> In distilling a past thread onto the wiki (still in
> progress), I' developed a grid that discusses how authors might
> describe the embedded content for the equivalent text contents in
> terms of the types Hixie previously identified: decorative, iconic,
> illustrative, and meaningful.[1]

Aside: you say there "Omitting an iconic text equivalent will make the page
virtually unusable", but it's not that simple. That applies only to cases
where the icon is the only indicator. In the case of <a href="/">home <img
src="home.png"></a>, the icon should *not* have alt text.

> The example you give here corresponds to what I called iconic
> description. The other two types of description I included in the
> grid are subject description and visual/aural description. The iconic
> description is a very common case, however, like decorative its one
> of those things that should likely be handled through CSS. In other
> words an author can provide a marker or another element with
> meaningful text.

Agreed. (Although currently CSS, allowing background images only, doesn't
make this easy enough IMO.)

> The particular icon that stands in for this marker
> is very much a styling issue. I'm not suggesting that we require this
> approach, but I think that the iconic embedded content has
> similarities to the decorative  embedded content.

It depends. In the case of <a href="/">home <img src="home.png"></a> the
image canbe considered decorative. Not so in the case of <a href="/"><img
src="home.png"></a>.

> For non-icons, it may be very much appropriate for an author to b
> begin with a generic description of the embedded content: though I
> think the distinction between subject description and visual/aural
> description is useful as well. In any event, the specific context of
> embedded content in a document may elicit the need for some
> alteration in the subject description or the author may be focussed
> on 'subject' in the embedded content  entirely different than an
> author describing the content without the present context. However,
> in either case we're talking about description.

Sorry, I just cannot folllow this. What's a "subject description"? There also
appear to be comma's missing or something. Or perhaps it's just me?

> Even the distinction
> between the subject description and the visual/aural description if a
> fine-line. For a particular audience,, the subject description "an
> aerial view of the Statue of Liberty" may be a sufficient subject
> description. However, for another audience that may have never seen
> the Statue of Liberty from any angle, further visual description may
> be necessary to convey the meaning relevant to the present context.

I'm not so sure I agree. As an author, *all* my content is authored with some
assumptions about the background knowledge of the audience. If I assume that
the audience knows what the Statue of Liberty looks like, then I'm probably
not using the image to convey what it looks like. So it doesn't seem logical
that an equivalent must suddenly convey what it looks like. Doing so would
make it an unequivalent.

> So I think we need to find a better way to communicate to authors of
> what kind of text makes a good equivalent than saying its not a
> description.

That's what I did, and you say you disagree, but it doesn't get clear to me
at all why you disagree. Especially because you then go on to talk a lot
about "descriptions" :)

> Even saying its not *just* a description doesn't really
> help. Whatever the equivalent text needed its a description of some
> kind.

Is it a description or isn't it?

> The issue is what kind of description is most suitable to the
> present context. Even the @title attribute is a description of sorts.

How? @title is for advisory information. At best I can see that it might
provide a description of the author's intent of something. But in that sense
*all* content is a description.


-- 
Sander Tekelenburg
The Web Repair Initiative: <http://webrepair.org/>

Received on Thursday, 6 September 2007 11:46:00 UTC