On 3/27/14, Peter Schaffter <pe...@schaffter.ca> wrote: > is it really necessary to scan an entire paragraph > to determine optimal linebreaks if judiciously adjusted word-and > letter-spacing on a line-by-line basis can produce similar results?
The more relevant question is, *can* a line-by-line algorithm, with some tweaks, produce results on par with an algorithm that considers the entire paragraph? I can't answer this definitively, of course, but I suspect it cannot. I think it's most likely to fail in the situation Werner cited: very narrow columns. That's the situation where, say, line 4 might have two possible breakpoints that both seem about equally good locally, but choosing one of them would make it possible to justify line 7 elegantly, while the other would give line 7 no good options. This is theoretical hand-waving at this point, but when you're dealing with lines that contain an average of three to four words, it's easy to see how a couple of long words in a row will create difficulties, and that the way the paragraph stands when those words are reached--based on decisions made previously--could affect whether those difficulties can be handled gracefully. Hopefully, someone will prove me completely wrong.
