Re: aria-describedat

On Thu, Mar 22, 2012 at 6:02 AM, Leif Halvard Silli
<xn--mlform-iua@xn--mlform-iua.no> wrote:
> * It follows an established pattern within ARIA. Why is
>  describedAT so 'special' that it needs a unique naming style?

Why is "role" so special?

>  A single attribute without the prefix only sounds confusing.

Maybe.

I was more thinking maybe we should just stop naming things using the
"aria-" prefix.

Perhaps the weirdness of the "aria-" prefix is a good argument for
just defining new features in host languages not expanding ARIA, or
for changing ARIA to recommend host languages implement ARIA
properties by providing implicit mappings from native names rather
than requiring them to include the ARIA names.

> * It learns from the antipattern set by @longdesc and @summary:
>  Unique names for seldom used/seen attributes is no good. Better
>  with an often seen prefix for a range of related attributes.

What gave some coherency to this ragbag collection was mainstream UAs
were using them only for accessibility API mappings. If we take the
view that user agents should be building UI on top of ARIA (like this
spec for @aria-describedat does), I don't think the ARIA attributes
are especially "related" to each other. They are just abrogating a
load of common document and application semantics into a single
vocabulary. Hey, where have I heard that before … oh yeah, that's what
HTML is supposed to do! Why don't we just add new common document and
application semantics to HTML and recommend other markup languages
reuse HTML features rather than reinventing the wheel?

> @aria-DESCRIBEDat, is not completely void of benefits:
>
> * It indicates relationship to aria-DESCRIBEDby

What relationship? The one where authors get utterly confused between
the two? ;)

> * The 'AT' part has hyperlink connotations - @.

Twitter? Email? If "hyperlink connotations" are good, then surely
"url" or "href" have stronger connotations.

> * Less technical than 'fooURL'

How is that better? If it's better, how about "descriptionlink"?

> and leads the thought towards the content that one points to

I think that's an implicit aspect of any name that suggests a hyperlink…

> and thus delegitimizes misuse.

No, it encourages misuse since it does not make the data type clear.

--
Benjamin Hawkes-Lewis

Received on Thursday, 22 March 2012 06:53:05 UTC