3 ms·

Large pdf and djvu files generally means low quality scans.

akiselev 4y ago
I think that really depends on the type of book. I've found the large PDFs to be by far the most reliable for books printed before the 20th century because - save for Google Books - no one's OCR is tuned for the kind of printing errors, age related wear and tear, and fonts used over the centuries. Manuscripts (which are admittedly very niche) often have to be high resolution scans or photographs just to be readable, but zlib and libgen don't really have many of those.
- zozbot234 4y ago
  Surely books published before the 20th century are in the public domain by now, though?
  - akiselev 4y ago
    They are but someone has to be in possession of the book and go through the trouble of digitizing it. Usually it’s museums, libraries, and private collections and a variety of enthusiasts eventually upload from those sources to Zlib/libgen. Since no one cares about copyright there anyway it can be more complete in some genres than other central databases.
dredmorbius 4y ago
That depends, variously on a number of factors. A straight-text book can run as little as 500--750 KiB for 100--200 pages or so. ePubs may run smaller as they're a compressed archive of HTML files, so how large the straight text is is largely a matter of the additional markup and stylesheets applied. I've generated PDFs from Markdown which are fairly comparable. Most mainstream trade press books run about 3--6 MiB as PDFs, if they have few images or graphics. Books with a heavy graphics content can easily swell to 30--300 MiB. And scanned rather than generated PDFs tend to run similarly sized. Many of those scans are, however, excellent quality. The largest ebook within my own collection for quite some time was a copy of Lyell's Geography, downloaded from the Internet Archive and based on a library scan of a 19th-century printing. On a colour tablet, it's actually pretty nice reading, on a B&W E-ink tablet, such scans often have a significant background ghosting which may or may not be eliminated by contrast adjustments in your reader. For newer materials, I've seen any number of issues with ebooks: - Clean single-page scans with OCR'd text for highlighting and copy/paste are generally my favourite. There's something about original print layouts that I still find preferable to ePubs and other digital-native formats. - Native PDF can also be quite good though they're far more prone to designeritis, where someone has thought they could improve over ink-on-paper conventions. If you think you can do this, you are almost certainly wrong. - ePubs rerendered as PDFs are among the worst options. I prefer PDFs generally, but not in this case. - ePubs ... can ... be reasonably good, though again the principle problem is excessive flexibility for designers. Less is more. I really miss having a single consistent layout of the text, rather than something which reflows as font faces, sizes, and spacing are changed. As a saving grace, cringeworthy (and migrane-inducing) font choices can be overridden with The One True Serif Font. - Various low-quality scans. 2-up, lots of page skew and placement variation, heavily-marked texts, and such. I'll still almost always prefer these to an ePub or other digital-native format, though the reading can be much harder. - Tiny-font 3-column scientific publications. Even on my 13" e-ink reader, these can be a challenge, and I'll occasionally resort to a sub-page rendering (this is natively supported in the Onyx NeoReader and several other e-book readers). Why such publications continue to insist on this format as we approach the 2nd quarter of the 21st century I've no idea at all. Scans based on older journal pubs (from ~1950s -- 1980s or so) with both physical wear and scan artefacts can be especially challenging.

I think that really depends on the type of book. I've found the large PDFs to be by far the most reliable for books printed before the 20th century because - save for Google Books - no one's OCR is tuned for the kind of printing errors, age related wear and tear, and fonts used over the centuries. Manuscripts (which are admittedly very niche) often have to be high resolution scans or photographs just to be readable, but zlib and libgen don't really have many of those.

zozbot234 4y ago
Surely books published before the 20th century are in the public domain by now, though?
- akiselev 4y ago
  They are but someone has to be in possession of the book and go through the trouble of digitizing it. Usually it’s museums, libraries, and private collections and a variety of enthusiasts eventually upload from those sources to Zlib/libgen. Since no one cares about copyright there anyway it can be more complete in some genres than other central databases.

dredmorbius 4y ago

That depends, variously on a number of factors. A straight-text book can run as little as 500--750 KiB for 100--200 pages or so. ePubs may run smaller as they're a compressed archive of HTML files, so how large the straight text is is largely a matter of the additional markup and stylesheets applied. I've generated PDFs from Markdown which are fairly comparable. Most mainstream trade press books run about 3--6 MiB as PDFs, if they have few images or graphics. Books with a heavy graphics content can easily swell to 30--300 MiB. And scanned rather than generated PDFs tend to run similarly sized. Many of those scans are, however, excellent quality. The largest ebook within my own collection for quite some time was a copy of Lyell's Geography, downloaded from the Internet Archive and based on a library scan of a 19th-century printing. On a colour tablet, it's actually pretty nice reading, on a B&W E-ink tablet, such scans often have a significant background ghosting which may or may not be eliminated by contrast adjustments in your reader. For newer materials, I've seen any number of issues with ebooks: - Clean single-page scans with OCR'd text for highlighting and copy/paste are generally my favourite. There's something about original print layouts that I still find preferable to ePubs and other digital-native formats. - Native PDF can also be quite good though they're far more prone to designeritis, where someone has thought they could improve over ink-on-paper conventions. If you think you can do this, you are almost certainly wrong. - ePubs rerendered as PDFs are among the worst options. I prefer PDFs generally, but not in this case. - ePubs ... can ... be reasonably good, though again the principle problem is excessive flexibility for designers. Less is more. I really miss having a single consistent layout of the text, rather than something which reflows as font faces, sizes, and spacing are changed. As a saving grace, cringeworthy (and migrane-inducing) font choices can be overridden with The One True Serif Font. - Various low-quality scans. 2-up, lots of page skew and placement variation, heavily-marked texts, and such. I'll still almost always prefer these to an ePub or other digital-native format, though the reading can be much harder. - Tiny-font 3-column scientific publications. Even on my 13" e-ink reader, these can be a challenge, and I'll occasionally resort to a sub-page rendering (this is natively supported in the Onyx NeoReader and several other e-book readers). Why such publications continue to insist on this format as we approach the 2nd quarter of the 21st century I've no idea at all. Scans based on older journal pubs (from ~1950s -- 1980s or so) with both physical wear and scan artefacts can be especially challenging.