I agree with this, as well as just using HTTP for it (Status: 300/Conneg).

Still gets a bit tricky when talking about a 'search results' or
'browse' page (rather than a 'resource' page).

Of course, some other metadata options beside JSON would probably
helpful in this case, since it would take some prior knowledge of the
JSON schema to know what you're looking at, otherwise.

-Ross.

On Sun, Dec 7, 2008 at 9:46 AM, Ed Summers <[log in to unmask]> wrote:
> On Fri, Dec 5, 2008 at 9:08 PM, Michael Ang <[log in to unmask]> wrote:
>> From what I've been reading it sounds like <abbr> with title is more of a
>> problem than <span> with title.  So maybe <span> with title isn't too much
>> of a problem in practice?  I suspect that whether the span is empty doesn't
>> make a difference *if* the screen-reader is set up to read the title
>> attribute of a span.  But having the software set that way seems unlikely(?)
>
> Yeah, the tooltip displaying an ugly serialized ContextObject is kind
> of annoying, but I imagine there are hacks around that. The main
> problem for screen readers, at least according to that BBC article,
> concerns the use of <abbr>.
>
> I didn't mean to stir up a huge debate. I just think libraries
> sometimes overlook use of the <link> element to link to alternate
> (more machine readable) representations of a web resource. With the
> notable exception of  auto-discovery for syndicated feeds (Atom, RSS,
> etc).
>
> //Ed
>