It's worth pointing out this is done in pure Javascript, and works by compiling PDF functions to equivalent Javascript functions which are then visible to Firefox's JIT. Despite being only around a year old, it still manages to render the majority of PDFs thrown at it (it's been my primary paper reader for the past 6 months or so).

As for missing features like some complex gradients, I can't say I've missed them, except on occasion when dealing with shiny PR materials. Earlier versions occasionally emitted blank pages, but these could always be skipped thanks to a side effect of the PDF format.

PDF.js has an amazing future for such a young project, and it is the foremost demonstration of exactly how complex programming tasks can be expressed using native web technologies. Turns out 35kLOC of Javascript almost completely subsumes the functionality of a behemoth native application (Adobe Reader) that on some machines would require half a minute just to 'boot'.

While Mozilla are pumping out stellar designs like this, Google are pushing crap like Native Client and their proprietary, binary-only Foxit Reader solution instead, complete with the hundreds of thousands of LOC of insecure C this entails. Rock on, Mozilla!