We currently have (from the bottom up):
- raster layers (HTML)
- canvas objects (canvas)
- vector layers (SVG/VML)
- vector object containers world space (SVG/VML)
- vector object containers screen space (SVG/VML)
- map widgets (HTML)
To support world-space markers based on html instead of SVG/VML, it is necessary to have an HTML container (div) that supports transformation. As it is not possible to mix SVG and HTML elements, we suggest to add an additional HTML container on top of the SVG containers, just before the map widgets. This container would handle translation during panning. Marker objects should support the Transformable interface and be able to render themselves transformed (for fixed size markers, this typically involves translating their anchor point).