[clangd-dev] textDocument/didOpen, file tracking, reading files from disk, etc.

[Ken Domino via clangd-dev]

> In addition to the resource management concerns, I would argue that 
> not requiring didOpen is a bug in the protocol and all the clients 
> should do this.

Hi Ilya,

I thought this too, and even opened a bug against MS's client. But, 
after thinking about this for a while, it is clear what the spec is 
trying to say here: "truth", which is obtained by the client with an 
open, is a content-wide write lock. If you look at it this way, 
something we all know very well, then you can see the range of 
possibilities here on what the protocol means and what it could evolve into.

Ken