There's a lot of research on the optimal line length <i>for reading</i> but there are a number of other considerations that go into website design. For example:<p>Many sites (including both Wikipedia and HN) practice responsive design, where the width & layout of screen elements adapts to the size of your window. There's a good argument that providing the user control of their own situation outweighs the consensus research on the "average user". After all, you're trying to make <i>each</i> user happy, you aren't trying to make a hypothetical average user happy.<p>Most of the time, users skim sites, they don't read them. Optimizing for reading is counterproductive when the user just wants to find a sentence or two and go away.<p>Sites can become aesthetically unpleasant if they have a small column of text and a gigantic amount of whitespace. Nobody reads the site when the user bounces.<p>Sometimes sites are set up to use indentation to convey hierarchy, eg. in comment threads. These can also look awkward if the right margin keeps shifting over with each new comment. (Although personally, I prefer that to having tiny little comments that are 2-3 words wide.)<p>There's a meta-point here: oftentimes, academic research into one specific area is not relevant to a larger product, because there are other trade-offs that have to be made. When I first went into software, I wondered why all this cool academic research into programming language design was being totally ignored. It turns out that oftentimes the really cool language features like macros, type inference, predicate dispatch, tail-call optimization, etc. have side-effects that make them impractical for use in a real software system.