I do agree that PDF optimization is useful and should be standard practice — pdfopt has been widely and freely available for ages as part of Ghostscript. Optimization can, however, lead to larger file sizes, and Google may have been trying to avoid that.

Text OCR would be nice, but there are projects for that already. Google isn’t into duplicating effort, generally. Their product is highly legible and gives one a sense of the effort that once went into fine typesetting: compare the Montaigne download to PDFPlanet’s Fables of Aesop, generated with Microsoft word, and you’ll see the difference between typesetting and bulk text chundering. The former is an image of a work of skill; the latter is the McDonald’s version. If Google were trying to put Project Gutenberg out of business, perhaps PDFPlanet’s criticism would be relevant — but they fail to realize that each project has a separate goal.

Speaking of beauty in typesetting, and having scrubbed my eyes from looking at that Aesop translation generated by Word, the comment PDFPlanet makes about the resolution of Google’s texts being so low for easy legibility is pure bunk. Blowing up the text so that “hoc est” fills my screen reveals jagged edges, but at a reasonable zoom level, the text passes for perfect. Perhaps PDFPlanet forgot to turn on antialiasing in their PDF renderer? Perhaps they engage in hyperbole to attack a major project in order to gain attention for their parent company’s commercial ventures (NitroPDF)? It has succeeded, as I for one had never heard of their site or their product until this post on Stoa.

It may not be everything, but it’s a pretty hefty wedge in the door, I think. And, frankly, given what I’ve found so far, who /needs/ anything written in the 20th century? : )