Hello guys, Recently I've been thinking on splitting Yi into core package and UIs.
Here are some of my thoughts: Currently Yi core slightly depends on frontends. This dependence can be easily removed(fix me if I'm wrong). However, it'll no longer be possible to select frontend on startup with command line arguments. I see 2 ways of dealing with this problem: - dynamically load selected frontend on startup (hs-plugins?). Personally I'm against it, since it'll do nothing but add extra dependencies. - migrate to client/server architecture. Instead of loading object file in runtime, Yi'll: - start listening socket - select UI depending on cmd args and user config - run UI client, pass it socket address - ... I find this soltuion much more flexible. Reasoning: - Before initiating this migration, some kind of protocol between UI and core has to be made. It'll limit core editor functions that can be accessed directly by UI. (Now frontend has to do a lot of stuff that core should handle(at least in my point of view). E.g. Yi has a nice handler for keyboard events, but mouse stuff is done mostly by UI. It sets mark point, moves cursor, scrolls buffer, etc. Moving these actions to Yi core'll simplify UIs and make their behaviour more consistent.) - It'll be possible to implement remote/multiple connections to one editor instance. Just like emacs --daemon mode. - smaller Yi executable, probably faster config reloading. GHC won't have to link GUIs into executable. I realize this transition'll require a lot of work, but benefits seem worth trying. I'll appreciate your opinion on this decoupling. Also, if there are no principal objections, I'll slowly start working on it. -- Yi development mailing list yi-devel@googlegroups.com http://groups.google.com/group/yi-devel
