Your typical eink device (a kindle) is 6 inches diagonal, but there are also 6.8, 7.8, and 9.7 in somewhat common use, and more uncommon. Three sizes of iPads, dozens of sizes of android tablets, phones and laptops, etc.
You can't make a single page size display well on all of this -- at 8.5 x 11, a page will be too small to read when shrunk down to fit a 6in screen. Bake it at 6in and the pages are ridiculous on larger screens, but still too large for phones. You need a format that shapes the number and size of pages to fit the display -- enter ePub, which is based on HTML.
In other words, ePub cares about where the chapters are and such, it is structured text, while PDF is a page description and what's on a single page will always be on a single page.
ePub is one of the best formats for blind people, they can read all contents of most books very easily. You can (like HTML) basically just dump out plain text by ignoring all the tags.
PDFs are the worst format that is commonly available, it isn't much better than a jpg of text. There are many programs that try to convert PDFs back into plain text, they all get confused by sufficiently complex PDFs. When you look inside a PDF, it is formatted as "put characters ABC at page location X,Y", you then have to try to extract the text from that by manually tracing the flow / columns / etc.
And as you want to do more and more complex things with letter spacing you end up with ever shorter flows of characters; there are certainly some tools that end up creating every character individually so even extracting a single word in Latin script is hard.
Sometimes you see a PDF where the end of a line of text means the end of a paragraph structure wise. And if you choose to use a bigger font, the text flow changes and sentences break halfway, leaving a half empty line, etc. When this happens once every two or three pages, not a problem, but if this happens all the time it makes the book unreadable. This happens quite a lot at the ending of a page as well.
If you have something like a sidenote or footnote, it is very likely that it breaks up the textflow, and not in a nice way, after a paragraph, but it breaks up a sentence. This can be very confusing, especially if you're reading in a different language with complex sentences.
Then there is another level of breaking up a sentence halfway, where words are not formatted as words, but as letters, and where each line breaks up words. I've seen this with a O'Reilly book. They fixed this within a week, so that was excellent service, but it shows how this can work.
Images can be a problem, depending on their size. Most ereaders are underpowered and can have difficulty processing large images.
Tables will be mess in 99% of the cases, because they consist of text and lines, and do not have any structure. It's always a surprise how they will show up. It could mean that half of the table is not even shown, or column one is shown, below that column 2, and you don't see the relation anymore. Sometimes column 2 is partly shown before column 1.
Some PDFs simply crash the ereader program.
I apologize for the language but I find no other words that work.
Both formats have evolved over the years to build on their strengths and shore up their weaknesses- for example, you can now embed structured content in a PDF to make it more ePub-like (and potentially reflow text in a viewer), but in practice few PDFs take advantage of this. I don't know nearly as much about ePub, but I would imagine that over the years it has evolved to allow for more control over the visual presentation of the content.
Ultimately PDF is a more universal format but ePub is a better fit for most digital books. Both formats will probably coexist for a long time.
In practice though this is still a shitshow even in ePub 3, 99% of publisher just ship a pile of endnotes, and consequently certain authors like Terry Pratchett are a god-awful experience in ebook format. Jumping back and forth every 30 seconds sucks.
There are work-around that can be done with JS or certain reader-specific markup you can use for iBooks to at least get pop up footnotes, but most publishing houses don't have the technical knowhow or desire to invest that much time in doing the right thing. There needs to be dedicated markup and a better standard for how readers should treat its display whenever possible.
A real failure of the standardization committee, and it will likely only get worse with a group that cares even less about books specifically. The standard continues to be driven by people more interested in gee-whiz features than solving longstanding problems in replicating table-stakes features of print.
Of course, being packaged also make it much easier to apply DRM (and, perhaps thankfully, means we don't actually need to address DRMing of HTML content directly which helps fight against it being in browsers).
ePub allows content to reflow in order to fit the screen size, while PDFs are a fixed size.
It's not typically a problem if you're reading the book on a desktop or laptop, but it becomes frustrating when you're on a tablet or mobile device.
Which is why ePub is basically HTML + some media formats + packaging.
But there’s some fierce objection against the merger of IDPF into the W3C [1][2], the key (?) argument being formulated as:
> “The W3C is focused on promoting the Web, but eBooks are not websites. When the IDPF is gone, who will advocate for readers?”
Maybe a fair point, but I don’t think I can agree. While it is true indeed that reading long-form content like books requires enduring focus, and that having to read them in a browser where linkbait always is luring to have you click away into a never ending feed of distraction, the problem is not the underlying technology.
[1] http://www.publishersweekly.com/pw/by-topic/digital/content-...
The actual development is entirely controlled by the browser vendors, causing pain for everyone trying to parse HTML programmatically.
What this leads to is basically fixing random, one off issues depending on publisher and book. I would definitely suggest not writing your own reader and instead looking at something like Readium (and contribute if you have time!).
[1] http://mademers.com/two-more-indesign-cs5-export-to-epub-bug...
