On Fri, 27 Aug 2010, Hugh Guiney wrote: > > I'm authoring an XHTML host document with namespaced inline SVG and > XLink. I have vector images that recur throughout the site. I'd like to > implement SVG's <defs> and <use> to reduce the file size of the document > and keep style separate from content, as with CSS. > > If I put an SVG tree with <defs> anywhere in the XHTML document, other > trees with <use xlink:href> will correctly reference it, as tested in > the latest public release Gecko, Webkit, and Opera. So the question > becomes, where do I put it? The most obvious answer seems to be <head>, > since, like CSS definitions, this is metadata being defined for use > elsewhere in the document. The only problem is, Validator.nu doesn't > like it: > > "Error: SVG element svg not allowed as child of XHTML element head in > this context. (Suppressing further errors from this subtree.)" > > Same error when ditching the root <svg> and including only <defs>, the > result of which still works in all but Opera. > > This error can be avoided if the <defs> tree is put in the XHTML <body>, > but then there is blank space the size of the defined shapes at the top > of the document in all 3 engines. A workaround is to give <svg> a @width > and @height both of 0. But leaving the definitions in the <body> when > they technically don't represent contextual content strikes me as > non-semantic. > > My proposition would be to simply spec a subset of SVG consisting only > of metadata elements as valid in HTML's <head> context. This could be > just <defs>—I'm unsure if there are any other elements that fit this > definition since I am relatively new to SVG; but in either case it'd aid > semantics and is already supported in today's SVG-capable browsers.
This is an interesting idea. I would recommend approaching the SVG working group and suggesting that they define the content model of <svg> and other SVG elements such that there's two ways to use it: one where <svg> is considered embedded content, and one where it's considered metadata content, with appropriate restrictions on the latter. With such a set of definitions in place, the HTML spec's model would just work (it already refers to the content model of <head> as just "metadata content", for instance). -- Ian Hickson U+1047E )\._.,--....,'``. fL http://ln.hixie.ch/ U+263A /, _.. \ _\ ;`._ ,. Things that are impossible just take longer. `._.-(,_..'--(,_..'`-.;.'
