You would think it would be possible to avoid the bugs by breaking a browser up into discrete components, and have it call curl each time. perhaps there is a speed penalty i do not know about.