Twisted's coding standard for our #ReStructuredText files is to use what we have termed "semantic newlines". That is to say: every sentence is on its own line. In principle, I like this. It means smaller, more readable diffs, and less pointless churn. In practice, with my #Emacs fill-paragraph muscle memory, it is a *constant* frustrating exercise where I'm tripping over myself and either constantly re-wrapping and then manually un-wrapping, or forgetting to wrap at all.