Just building the web-based editor that you want this year may be fine for this year, but I am not convinced we should be putting much effort into designing systems that don't learn from their predecessors. The classic editors can feel rickety today because they were made with the TTY, and arguably only the TTY, in mind. A system designed today should keep that in mind, and I think make future-proofing it's first and for-most concern.
I think it is also really hard to justify supporting fewer features than classical editors do today. Removing the ability to implement technical features so that you can add support for UI features seems like a step backward to me. I don't think there is much (or anything) that you couldn't add to Vim or Emacs, the question is just more along the lines of "can you make it look how you want it too". An editor designed with UI agnosticy (is that a word?) as a first class feature should not have this problem.
I agree.
That would be "Agnosticism", I think.
Your comment seems weird to me. It sounds like "Use an underpowered tool for your everyday task but turn to a vastly more powerful tool only in some strictly defined situation."
Wouldn't it be a little more sane to go the other way: use a super powerful tool for your daily work and only turn to a far less powerful tool when there's no other choice?