The Google pdf viewer seems good enough to me. It'll be interesting to see what you can improve on that.
I do see the value for formats like .pptx where I may not have a reader, but PDF is already a "lowest common denominator" like HTML.
Sure, I like reading my PDFs with evince, but sometimes HTML is just preferrable.
I've always thought that the best solution would be for browsers to just include PDF rendering support alongside HTML. PDF is a popular enough format on the web that it should really be handled by the browser as a core feature, not by some add-on or external application. I really liked Apple's Safari for this reason (although I dumped it for Firefox because of AdBlockPlus).
Done right it would seem like perhaps some of the higher-level rendering layers of the browser could be reused, regardless of whether the underlying content was PDF or HTML. (In fact isn't this how Safari works, with Quartz?)
No offense to Scribd, but I'd like to see the need for their service go away by in-browser support for progressively-downloaded PDFs.
Removing yet another plugin from your browser shrinks the attack surface, and as a side benefit, reduces use of one of the worst IMO bits of web tech.
PDF -> Internal Representation -> Flash viewer
In principle they should only need to change to
PDF -> Internal Representation -> HTML5 viewer
That alone should make a huge different in terms of time needed.
Never having done something similar myself, I would expect the hardest part to be correctly parsing the original PDFs and dealing with the varyiaty of PDF generators out there.
To me, this feels like desperation. It seems a little late to stop depending solely on Flash. Also, why not say that you're moving after you partially move? The companies I admire the most brag about distant features the least.