It is important for LCD text and for keeping people happy that we have native rasterization available. There are some cases where our current approach is best and some cases where a native approach is best. Nether approach can solve all use cases. So the best solution is let the developer choose on a node by node basis which renderer to use.

When Native is best 

 *  Reading large chunks of text, native renderers are very carefully tuned with platform fonts to be good to read (Minimize eye strain etc)
 *  Consistency with other naive applications
 *  LCD text as hardware and user preference knowledge is needed that platform renderers have.

When our cross platform renderer is best

 *  We want a true to the type (Some native renderers distort type, eg windows on some fonts)
 *  Consistent cross platform rendering of text
 *  Animation, smoothly animating text even on sub-pixel boundaries