Re: [SVGMobile12] more on data types

Hi Ian,

www-svg-request@w3.org wrote on 01/11/2006 03:39:02 PM:

> 
> On Wed, 11 Jan 2006 thomas.deweese@kodak.com wrote:
> > 
> > It means that no compliant renderer can inform users that 
> > they have made a mistake because of course they haven't they 
> > have simply indicated to the UA to ignore the value of the attribute.
> 
> This is not true. There is nothing stopping a UA from reporting what 
> values are non-compliant ("unsupported" in SVG 1.2 terminology). Just 
> because the spec says how to handle it doesn't mean it is allowed.

   I think this is a fairly creative reading of the specification,
the spec says the UA "must ignore".  I would posit that popping up an
error dialog is pretty far from 'ignoring' the unsupported 
element/attribute.  They could soften some of the language to indicate
that the 'ignore' only applies to the visible display/rendering.

> It is easy to confuse _UA conformance requirements_ with _authoring 
> conformance requirements_. The two are separate and distinct. 

   Sure and this is clearly written as a UA conformance requirement:
        http://www.w3.org/TR/SVGMobile12/implnote.html#UnsupportedProps

> Something can be invalid while still having interoperable behavior.

   This is possible but is not IMHO how the SVG spec is currently written. 
 
If the WG wants to adopt your interpretation then I would suggest making
it clearer that the 'ignore' only applies to the rendering and adding
a fifth bullet to section 'C.2':

        * The UA MAY display a highly visible indication that an 
unsupported
          element, attribute, property, attribute value or property value 
was
          encountered.

   However, given Robin's responses on this topic I doubt the WG would
agree with your interpretation.

> For example, in CSS, this:
> 
>    p { color: green; color: 1px; }
> 
> ...is invalid. The UA requirements, however, are unambiguous: UAs must 
> treat this exactly as if it was:
> 
>    p { color: green }
> 
> They may, of course, report the error to the user as well (and indeed, 
> Firefox does).

Received on Thursday, 12 January 2006 13:27:44 UTC