Thanks for clarifying and forgive me if I used an incorrect terminology. As you said, however, the term hook is used in a variety of ways de facto, and I decided to use this word just because the way WordPress (and other CMSes) use it, it's basically my goal. I think I might have incorrectly explained my goal. Instead of adding on a comment, I'd be grateful if you'd check out my Question Edit.

Thanks for the answer, it was a nice read; however, this deals mostly with the first part of my question: event-systems as hooks. The thing with events is that the code that triggers them rightly doesn't care of who is subscribed; as such, events don't return values that can be passed around and manipulated. In my case, I need a way for other parts of the components to "filter" around specific data exposed by a specific component. As far as I know, this is specifically not allowed inRedux for good reasons. That's why I was also asking for an alternative, since this may be an anti-pattern.

Also, how would you go and approach the fact of binding the comment to an offset-range instead of a precise point? Last but not least: the data allocation table and transactional update queue seem amazing concepts. I did some research about the topics, but could you elaborate a bit on how you would go and implement those concepts in this architecture problem?

First of all, I must say that I liked the enthusiasm in the answer, it was great to hear such good feedback. As for the answer itself, I must say that I would be against same syntax for opening and closing the tags; and perhaps, to avoid the XML problem I described above in my most recent update, I would specify what's being opened and what's being closed in the tag itself; perhaps like so: >>>>>(#1) Lorem ipsum (#1)>>>>>>. Also, it seems like your approach in intext comments would have them bind to a certain fixed position, how would that work if the offset is moved?

What is in your opinion the pro of not adding a closure tag but working with offsets? Isn't it too risky? What if I add a word between nonummy and nibh, wouldn't it mess up with my offsets?

I think a problem is that, as far as I know, in XML any element which happens to be inside another is considered child of the element, and overlapping of tags is not possible (i.e., you have to close the children before closing the parent). In my case there is no such hierarchical structure, as two notes can certainly overlap (added example at the end of my answer).

I would say that right now my practical concert is over a static and unchanging text. However, thinking ahead, I would most possibly like an architecture in which I am not bound by this decision, and implementing the changing of the text would not need a complete rewrite (of course it might need loads of work anyways). The metadata should stay for the revisions, but should stay reasonably glued to the position in the text (I would say to particular words and sentences).

I elaborated a bit on this in my new edit.

What other detail do you reckon are missing from my question to give you a clear picture?