Okay so what I want is where there is only a single thing that can be
displayed on the screen regardless of the application. This simplifies
things although it kind of messes up task switching but that is a
something that can be solved by blocking things. In any case, I think it
would be best if all the local LCDUI stuff was directly mapped to an index
on the server end. The local end would need to remember all the stuff that
is mapped to indexes for remote widgets. And the logic for most of everything
will be in the server including stuff like
setCurrent for displays. The
only thing is that the server side display will have a current for each
task but it will only have an active task. Any task which wants to set
current will do it but it will not have the intended effect until that
application is switched to. This means I do not have to worry about or
implement conditions where multiple tasks are running and they all want to
Okay, I am going to need a slight refactor. The first thing I want to do is to remove the locks on the display code so that they are never used or passed. Second I want to make it so that there is a GUI thread which does things. At least for Swing this is how it works because the AWT it is on top of is really not thread safe at all. Pretty much every modern and classic graphical thing is not thread safe either. So I will need a slight redesign and simplification. Also I should aim to reduce the number of arguments passed to displayables. But at least on the server end there will be a primary display loop, system calls will call into them and then wait for a response as required.
The main problem though will be calling the event responders on the client. When those happen they will definitely be outside of the thread. I suppose I will need a callback of sorts that allows the kernel to execute the client to do things, at least that is the only way I can think of things. It would simplify the design and allow for event handling though if I do support a callback. I should probably support it in the code anyway because it can be important to do things.
It would probably be easiest for event handling if there were just methods used for pushing to an event thread for eventual handling, that way it can match the Swing model which should work for every GUI framework.