> Have you looked at Phil Lord's lentic package? I think it implements a
> lot of what you're talking about.
This is nice to see!
Indeed, except for embedding, there is a large overlap with what I described as buffer lenses.
BTW, judging by this description: "changes percolation now happens incrementally, so only those parts of the buffer are updated. As a result, lentic now cope with long files with little noticable delay", the buffers don't share any data and need to sync with the master [linked] buffer.
Is this the best solution? I have imagined that at the low level there is an actual data structure that keeps the raw textual data and it could be directly shared by multiple buffers. I mean, when a buffer is saved to a file, the text doesn't need to be stripped of properties beforehand, right?