Quoting Alan Horkan <horkana@...44...>:
Cascading style sheets had me wondering if a Layer could be expressed as a class of groups? Something like: <group id="Layer_1" class="layer"></group>
That would be a very poor use of CSS classes; among other things, it pollutes the CSS class namespace. It's not like there are any standard CSS attributes to associate with the desired behavior for layers, so it would also necessarily amount to a nonstandard extension of CSS. Unlike XML, CSS has no real "namespaces" facility for application-specific annotations.
There is no disadvantage to using attributes in a custom namespace to make non-presentational annotations to a conforming SVG document; that sort of thing is the express purpose for which the Namespaces in XML standard was introduced.
In fact, the alternative -- trying to co-opt "standard" facilities into representing things that the standard doesn't specify (for example, using a "magic" CSS class to denote editing behavior) -- frequently has very bad long-term consequences.
Trust me, we've thought this stuff through already. The only true issue here is whether or not a particular existing sodipodi-namespace (or even inkscape-namespace) attribute is still useful.
As far as I can tell, sodipodi:docbase can probably go (we need to keep the C++ field, but there's no reason to save it to the XML), but bulia is still in the process of verifying that before we actually remove it.
-mental