Donald Knuth's typesetting system, TeX, is wonderful. Although it's more than thirty years old, its hyphenation, word and paragraph spacing, kerning, and mathematical typesetting are still vastly better than any alternative. It is the standard for typesetting in academic computer science, mathematics, physics, and many other fields. Knuth's books on TeX, typeset in TeX, are a delightful read, one of the best examples of writing about programs that I know.
Sadly, though, the time has come for us to replace TeX. There are two main reasons.
First, the output of TeX is not suitable for display on modern electronic reading devices. On a large screen portrait-oriented device like an iPad, it can be passable, but it is quite poor on a landscape-oriented device like a laptop, and disastrous on a mobile phone. Reading just one page of a typical two-column academic paper on a laptop means scrolling down the first column, up to the top of the second, and down again, since a full page does not fit on a single screen at a reasonable magnification. On a mobile phone, you can add scrolling back and forth to reveal the left and right of each line of text, since the typical line is wider than the screen once you have zoomed in close enough to read it. A modern electronic document needs to permit reformatting for the viewing device, and this is in direct conflict with TeX's fine control over margins, spacing, line lengths, and so on. True, there are ways to get TeX to produce HTML, but that just means that you are dealing with all of the details of spacing without getting the benefit of complete control.
Second, for the most part TeX is no longer used to create beautiful documents. Instead, it is mainly used to create ugly documents for the academic publishing industry. The goal of the industry is profit, not beauty; the TeX style files of publishers are designed to render an article into the smallest number of pages, to save money. Consequently, they use too-small typefaces, too-small margins, and too-small leading, among other aesthetic flaws. It's infuriating to have to take research that I'm proud of and force it into such an ugly form.
I realize that this second problem—ugly typesetting—is not the fault of TeX. I would rather replace the academic publishing industry than TeX, for many reasons. Some progress has been made; Knuth's own 2003 letter on the topic, linked and discussed here, is a remarkable example. And yet, the industry continues to enjoy obscene profits.
I believe that the way to break the hold of the industry is for academia to abandon paper (and hence, TeX). Paper is an essential element of the publishing industry. Paper, in the form of journals and conference proceedings, is the only tangible contribution that the industry can point to; everything else, from doing the research, writing it up, ensuring quality via peer review, editing, and, of course, typesetting, is done by academia. Abandoning paper thus eliminates the only remaining justification that the industry has for its existence.
Of course, electronic document formats can also be used and abused by publishers, but the reality is that academic publishers are not so nimble. I don't think they're capable of transitioning to paperless publishing. That means that the transition gives us a unique opportunity to build a better publishing model. Let's take it.
UPDATE: see here for a way to reflow two-column PDF.