But to make this debate a little more productive, we should think about what the core problem is. It's that many people feel a need for monolithic page structured documents that always look the same (aka. document fidelity).
If we have learned anything from the web it's that data (semi structured in this case) and presentation aspects should be separated. So what we need is a file format that includes both, similar to an HTML file with an embedded stylesheet, only with much more sophisticated visual primitives.
I've read the PDF spec top to bottom. It can do that and a whole lot more, but its design is anitquated in my view. It's bloated beyond belief and encourages incredibly bad documents to be created. Documents that don't even include the the text proper as sequence of unicode codepoints. Documents from which you cannot deterministically derive the sequence of words it contains.
I know Microsoft is very unpopular here, but their XPS format is a much better solution to the problem of document fidelity. If nothing else I hope XPS at least forces Adobe to come up with a radically redesigned PDF format and a less bloated and more stable reader plugin.
All of this does not solve the UI issues caused by embedding these documents into HTML. In my view, the best solution is simply not to do it. Let's keep them separate for download by people who have a need for it. If Scribd wants to be part of the solution instead of the nuissance they are, they should translate PDFs to HTML, just as google does, only better.
Seems like the platform is viral as such. Not viral as in alpha geek, but viral as in HIV.