I think the author is a bit out of their depth here. A linear feed isn’t quite what the author seems to be modelling in their view of the problems they observe and describe. A linear feed has a beginning and an end. You can (ideally client-side) sort it by Date, or by Subject like we do with our Twtxt clients. A Tree-structure isn’t what the author thinks either, this is more the structure that forms after you introducing some kind of “threading model”. The main problem with any kind of information system that tries to figure out algorithmically what you want to “see” is that type of interface has no start and no end. SO you end up with a “scroll of doom”.