Re: Applicable elements for predefined classnames

Olivier GENDRIN wrote:
> On 5/9/07, Rene Saarsoo <> wrote:
>> 1. If authors use their own arbitrary values with @class,
>>     then this is absolutely correct use of @class.
>> 2. When authors use their own arbitrary values with @role,
>>     then this will NOT be the correct use of @role.

   It's not sufficient enough to tell people not to do something. You
have to define error handling for when they break the rules. The least
author-hostile way of handling arbitrary role names is to ignore them.

>> Similarly authors can make up their own element <foo>, which
>> might be assigned a meaning in some future spec of HTML.
>> But usually there is no benefit in making up your own elements,
>> and people rarely do it. Similarly do they rarely come up with new
>> values for other attributes with predefined sets of values.
>> Why should it be the case with @role?

   I don't think that's the case. While people don't commonly invent new
elements, they use all sorts of arbitrary values for the |rel|
attribute, especially for microformats.

> I [totally] agree [Rene]: the former HTML specs never warned authors about
> predefined classes.

   How can a previous HTML version warn about using class names defined
in a later HTML spec? How do you warn against using a class name that
hasn't been invented yet?

> But HTML 5 spec will have to include a big red
> shinny warn about misuse of role (and some other property), and will
> also have to be very clear about the way it's used and the way to add
> a role value to the spec.

   This assumes that web authors will even read the HTML5 spec.
Considering how few have read the spec for HTML 4.01, that's hardly a given.

Received on Thursday, 10 May 2007 03:30:29 UTC