W3C home > Mailing lists > Public > public-html@w3.org > August 2008

Re: HTML5 object element - classid or not classid?

From: Dr. Olaf Hoffmann <Dr.O.Hoffmann@gmx.de>
Date: Mon, 4 Aug 2008 17:33:19 +0200
To: Boris Zbarsky <bzbarsky@mit.edu>, public-html@w3.org
Message-Id: <200808041733.19846.Dr.O.Hoffmann@gmx.de>

Boris Zbarsky:
> Dr. Olaf Hoffmann wrote:
> > For object currently these element specific attributes are
> > defined:
>
> This is the section that describes what documents are conformant to
> HTML5, yes?
>

It is simply the referenced section of the draft:
http://www.w3.org/html/wg/html5/#the-object

It mainly defines what the object element is.
The complete prose belongs to this section including
explanations or definitions of attributes available
for object like data, type etc, therefore the prose
seems to be required for authors to understand the
purpose of the attributes and the functionality of the
object element.

> > Later in the text some attribute 'classid' is mentioned:
>
> In the section that describes how UAs are to handle <object>, right?

No, I think this part explains the functionality of the attributes too,
because this part contains the definitions of the attributes too.
And if there is no definition of 'classid' this makes no sense for
new implementors too, if they do not know, what it is. Or is it
expected, that new HTML5 implementors have to implement first
all HTML versions from 1 to 4 before they start with 5?
And obviously authors have to know, how the attributes are 
interpreted by user agents, if they want to use object, therefore
they have to read it, even if it is not easy to understand.

>
> How is this different from any other case where there are authoring
> conformance requirements?  In all such cases UAs need to know what to do
> with web pages that violate said requirements (in this case by adding a
> 'classid' attribute).

Well, then it is missing in the object section what to do if a 'declare' 
attribute is present or a 'schubidu' attribute. If 'classid' is assumed
to be outdated nonsense, there is no need to mention it at all, 
especially not without a defintion, what it is.


>
> > If there is an attribute 'classid', it should be obviously
> > defined and the purpose of this additional attribute should
> > be explained to the audience including an explanation
> > of its possible values or are this the same values as for
> > 'data' as in HTML4?
>
> This attribute was present in HTML4 and its behavior only needs to be
> defined to handle legacy content.  

For whatever purpose, but currently I did not find a definition in HTML5
and because it is mentioned in the object element definition section,
authors may want to use it. If this is not intended for authors, it
should not be mentioned in this section at all or only with an explicite
indication, that this is only relevant for implementors.


> Authors of HTML5 documents must not 
> (in RFC terms) use it if they want to claim conformance.  UAs must (in
> RFC terms) do certain things with it to claim conformance.

This should be noted in the definition of classid, if this is the case.
However if it is mentioned, how this has to be interpreted by the
user agent in a section relevant for authors to understand how to
use object, some of them will use it anyway, therefore it is indeed
not a good idea to mention it at all in this area.
This is a question of good separation. If everything is mixed in
the draft, content authors have to read, it cannot be a surprise,
if they generate an even worse 'tag-soup' than today and
the HTML5 draft will never loose the stigma to define only
tag-soup, want I read quite often now ... 

>
> > (by the way, why is 'declare' missing, this sounds
> > pretty helpful to start objects without the need of
> > scripts or internal controls in the referenced object?)
>
> Probably because no one actually implements it?

Why not? 
It could have avoided a lot of disfunctional nonsense currently
arround in the world wide tag soup web to start object content 
somehow, causing accessibility problems with user-sided scripts 
for example.
Some declarative method to start/stop/pause object/audio/video
is currently missing anyway in the draft, I think, therefore
without 'declare' it is a regression compared to HTML4 
(the controls attribute in HTML5 is a nice idea, but is it available 
for object too?)

>
> -Boris

Olaf
Received on Monday, 4 August 2008 15:45:26 UTC

This archive was generated by hypermail 2.3.1 : Monday, 29 September 2014 09:38:57 UTC