Re: SVG12: svg in foreignObject

On Sunday 2005-08-21 18:43 +0200, Chris Lilley wrote:
> On Sunday, August 21, 2005, 6:28:24 PM, L. David Baron wrote:
> LDB> On Sunday 2005-08-21 18:20 +0200, Chris Lilley wrote:
> >> On Sunday, April 24, 2005, 7:27:38 PM, Bjoern wrote:
> >> BH>   From http://www.w3.org/TR/2005/WD-SVGMobile12-20050413/extend.html
> >> BH> section 19.2
> >> 
> >> BH> [...]
> >> BH>   The contents of 'foreignObject' are assumed to be from a different
> >> BH>   namespace. Any SVG elements within a 'foreignObject' will not be
> >> BH>   drawn, except in the situation where a properly defined SVG
> >> BH>   subdocument with a proper xmlns (see "Namespaces in XML 1.1" [XML-NS])

> >> It is prohibited by the schema, you are correct that this is not
> >> desired. The spec has been altered by removing the text after 'drawn' so
> >> that the sentence ends there.
> 
> LDB> Shouldn't it instead say that any SVG element *children* of a
> LDB> 'foreignObject' will not be drawn?  That seems like what may have been
> LDB> intended in the first place.  The revised text appears to prohibit
> LDB> drawing SVG inside elements of some other language that are inside
> LDB> foreignObject.
> 
> Yes, the original intent was that SVG foreignObject hands off to another
> renderer, and that if that content then hands back to a (second instance
> of an) SVG renderer, such nested embedding can occur.

The idea that different namespaces are handled by different renderers is
an implementation detail.  It's not mentioned in the spec, nor should it
be.  (And unless the model were explicitly the model described in the
spec, an implementation that drew the nested content on that basis would
still, as a whole, be nonconformant, since the wording above forbids
drawing it.)

> But the original wording seemed to allow direct SVG children. Can you
> suggest better text to clarify this?

Using what I said above:

   The contents of 'foreignObject' are assumed to be from a different
   namespace. Any SVG element children within a 'foreignObject' must not
   be drawn.

> The current wording makes it clear that the existing renderer is not to
> render any SVG content which occurs as either direct children or as
> nested children of foreignObject.

The specification should rely on the recursive nature of the nesting and
just define the recursive step instead of poking across multiple layers
of nesting.  In other words, it should just say that the content in the
other namespace is rendered according to that specification's rules.  If
that then recursively includes SVG, the drawing of the included SVG is
defined (or not) by that other specification.  The specification for
svg:foreignObject should only describe the handling of foreignObject's
children, and leave the rest to the specifications governing those
children.

-David

-- 
L. David Baron                                <URL: http://dbaron.org/ >
          Technical Lead, Layout & CSS, The Mozilla Foundation