Re: ISSUE-53: mediatypereg - suggest closing on 2009-09-03

On Wed, Sep 2, 2009 at 2:55 AM, Henri Sivonen<hsivonen@iki.fi> wrote:
> On Sep 1, 2009, at 18:31, Mark Baker wrote:
>
>> FWIW, I had a look through the obsolete list last night and found some
>> which had properly defined semantics.  For example, "plaintext" was
>> defined to mean the same thing as "pre".  That's good, and what we
>> should strive for for the examples given by Julian.
>
>
> If that level of definition is sufficient, isn't this on the same level for
> scheme:
> "User agents may treat the scheme content attribute on the meta element as
> an extension of the element's name content attribute when processing a meta
> element with a name attribute whose value is one that the user agent
> recognises as supporting the scheme attribute.
>
> User agents are encouraged to ignore the scheme attribute and instead
> process the value given to the metadata name as if it had been specified for
> each expected value of the scheme attribute."

I am personally not familiar with the purpose or history of the scheme
attribute so I'll defer that question to somebody else.  But it does
seem a bit behaviour-centric.  More below ...

> If plaintext *means* the same thing as pre given this statement about UA
> *behavior*, "User agents must treat plaintext elements in a manner
> equivalent to pre elements.", isn't the *behavioral* statement quoted above
> enough to define @scheme to *mean* nothing (i.e. it's a talisman)?

No, behavioural statements are irrelevant to language definitions and
therefore media type registrations, because behaviour can vary
dramatically between different classes of user agent.  The only thing
that matters is semantics.

I was just reminded of Mike's language draft;

http://dev.w3.org/html5/markup/

IMO, something like that, once complete (and including the HTML 5
parser), would make a more suitable reference for the media type
registration than the HTML 5 spec for these reasons.

> Also, doesn't the following "define" @profile on at least the same level as
> plaintext is defined?
>
> "When the attribute would be used as a globally unique name, the user agent
> should instead always assume that all known profiles apply to all pages, and
> should therefore apply the conventions of all known metadata profiles to the
> document.
>
> When the attribute's value would be handled as a URL and dereferenced, the
> user agent may resolve the attribute's value, and if that is successful, may
> then fetch the resulting absolute URL and apply the appropriate processing."

That's also very behaviour-centric, so no, it wouldn't make a good
definition IMO.  But I'm also not that familiar with the use of
@profile in the wild (outside of it's use by GRDDL and (non-use) by
microformats) so am not the best person to provide a better
definition.

Mark.

Received on Wednesday, 2 September 2009 16:19:09 UTC