Pleased to see that LSP world domination is on-going. Having implemented an LSP server myself I could rant at length about its flaws but none of them really matter because the fact that you can fire up almost any editor and get something close to full IDE support for a whole range of languages is fantastic.
(Which reminds me, I must try Emacs’ C++ LSP support again - last time round it had some rough edges & I went back to the dedicated clang based setup built around irony-mod.)
I’ve personally never had good experiences with LSPs; they feel tacked on, and that’s if they even feel like working properly.
IMHO, the only way to do editor intelligence is in the editor.
Sure, in a perfect world we would all use editors with an infinite supply of programmer time to provide deep integration with every programming language we wanted to use.
Back in the real world, the LSP means that a new language can get “good enough” support (good enough is: has “go to definition”, “provide completions at point”, “outline view”, “show documentation”) on every editor that supports the protocol, by simply implementing a single LSP daemon.
Does it match the deep integration + support that comes from using (e.g.) NetBrains Java editors? No. Is it /hugely/ better than the pre-LSP status quo? Definitely.