[uf-new] Use of img in rel-* (with analyzed data)

[Manu Sporny]

Tantek Çelik wrote:
> On 7/15/07 11:09 AM, "Manu Sporny" <msporny at digitalbazaar.com> wrote:
>> Tantek Çelik raised the
>> point that web authors often mis-use the ALT attribute[2].
> 
> To be clear, the conclusion from this is that publishers should be given the
> detailed *choice* of whether or not the alt text in their pages is included
> in microformats property values (rather than being forced to by *always*
> using it in contained properties).

Tantek, I don't quite follow the logic here. Publishers aren't given the
option on whether or not their ALT text shows up in a text-based
browser. They are also not given the option on whether their ALT text is
read out loud when using a screen reader.

Why, then, are we giving them the option on how ALT will be handled with
regards to Microformats? Or rather, why are we giving them the option to
hide data?

> Thus the alt (or src for that matter) attribute of an <img> element is
> *only* included on a property value if the property is set directly on the
> <img> OR via a class="value" construct.

You don't have the option of setting "rel-*" properties on images. That
is the whole point of this discussion. Your "just set it on the <img>
element" argument doesn't work for "rel-*". rel-* always go on anchor
elements (<a>).

As for class="value", that is a potential solution... thank you for
identifying it. However, I ask again - why are we giving publishers the
choice of violating the HTML specification? Of hiding data? Where are
the real world examples of why we need to provide that option?

> Our experience with this in practice has been quite good, and in fact, this
> is the first that *anyone* has raised any issues with it (in over two years
> of it functioning this way - that is it's not that no one's written it down
> yet - unlike some of the existing issues), so given experience to date, I
> would assert that we have the 80/20 (or far more than even) case covered

Since you are asserting that the community has 80/20, could you please
provide some data to back up that claim? How many people use images
inside hCard/hCalendar/hAtom and hResume? How many of those people have
@alt specified correctly? Incorrectly? How many examples of images used
in rel-* do we have?

We have collected quite a bit of data (and continue to do so) that shows
that mis-use of @alt isn't as wide-spread as previously asserted. In
fact, it falls quite short of the Microformat community's 80/20 rule. If
I wasn't clear about that previously, here's a re-cap:

As of right now, it looks as though roughly 80-90% of websites are using
@alt correctly, either by not specifying a value or by specifying valid
data in the attribute.

If you'd like me to demonstrate that figure further, I would be more
than happy to do so - using hard data that is available to everybody on
this mailing list.

-- manu