[uf-discuss] Hcalendar in bbc.co.uk/programmes

[Angus McIntyre]

On Thu, December 13, 2007 12:40 pm, Robert O'Rourke wrote:
> ... Could it be argued then that the machine-readable date-time
> belongs in CLASS on a SPAN?

It looks to me as if ISO-8601 dates would not be valid classnames, due to
the presence of colons and '+' signs. However, I may be wrong about that.

Even if that isn't the case, I have an aesthetic resistance to using
"class" for something that very clearly isn't a class. That seems to me
even uglier than using the "title" attribute of "abbr". I'd also wonder
about user-agent efficiency: a page with 100 dates would send the agent
into its in-memory representation of the stylesheet 100 times to look for
a class definition that it will never find. That feels wrong, although I'd
need to show that it actually exacted a noticeable performance penalty on
some platform for it to be taken seriously as an objection.

Using the 'id' of the span is probably out too: if you have simultaneous
events, you get non-unique ids in your page.

>> Given that the set of data types is potentially infinite, I'd like to
>> see an extensible method for hiding machine-readable data in the next
>> version of HTML.

A 'data' attribute or equivalent might be nice to have.

Unfortunately, we can't turn the problem on its head and, instead of
looking for exploitable attributes in the current spec, try to find
something that would be both machine-parseable and human-friendly to put
in our "title" attributes. While you might be able to come up with
something in one language that is easy to parse and sounds sane when read
out by a screen reader, you'll come unstuck the first time your parser has
to read a date in a language it doesn't know.

"title" on "abbr" (or "span") looks like the least bad of a set of bad
choices.

Angus