Most off the shelf VCSes are designed primarily for code — and line-based code at that. Look at the revision tracking support in Word or Google Docs — it's char

Most off the shelf VCSes are designed primarily for code — and line-based code at that. Look at the revision tracking support in Word or Google Docs — it's character based. That's what we really need some decent tool and hosting for, with stable URLs, that can be relied on. Arguably Word is better than Google Docs for this, as its revision tracking ultimately goes into one of the better defined parts of OOXML. (Last I looked LibreOffice was pretty poor for this, but I haven't looked all too often.) Still, I'm unaware of any VCS that actually does half of what many of people want to do — in diachronic linguistics you often have to deal with variation in source texts that /isn't/ semantically important (Romeo & Juliet v. Romeo & Iuliet, for example), and then doing n-way diffs between various copies of the "same" text. You don't want every single i/j or u/v to be flagged up, because the signal:noise ratio would be astronomical. Lots of different groups have lots of very subtly different requirements, and basically nothing off-the-shelf works well.