I've been wanting to provide a high-fidelity many-to-many commenting system inside of a text editor or browser since I was in college. My thought was that if you could annotate something as complex as Shakespeare:
http://imgur.com/FgsyAco
then you could annotate legal documents, code, and other high-density texts as well.
I've long felt that existing solutions fall down in a few ways:
1. UX -- this is a HARD UX problem because you are potentially managing a lot of information on screen at once. Anybody staring at a blizzard of comments in Word or Acrobat knows how bad this can get.
2. One-to-one -- Most existing exegesis solutions like genius.com only let you mark of one portion of text for threaded commentary, which is not ideal because complex text like the above example can have multiple patterns working in it at the same time:
http://imgur.com/x6BKKQW
(a crude attempt to map assonance and consonance)
Really, what a robust commentary system needs is to map many comments to many units of text, so that the same portion of text can be annotated multiply (as this solution attempts) but also so that the same comment can be used to describe multiple portions of text as well.
3. Relationships between comments -- It's great that this solution gives threaded comments as a first-class feature, but you also want to be able to group comments together in arbitrary ways and be able to show and hide them. In my examples above, there are two systems at work: the ideational similarities between words, and the patterns of assonance / consonance. You could also add additional systems on top of this: glossing what words or phrases mean (and in Shakespeare, these are often multiple), or providing meta-commentary on textual content relative to other content, or even social commentary on the commentaries. You need a way to manage hierarchies or groups of content to do this effectively. No existing solution that I am aware of attempts this.
I literally just hired somebody yesterday to start work on a text editor that attempts to resolve some of this, but it's an exceedingly hard problem to solve with technology.