In-client echo of input methods?

[Bill Spitzak]

Does the proposed input methods support clients being able to show 
intermediate results? This is certainly useful to make "dead keys" less 
"dead", and I am under the impression that it is very useful for some 
Asian entry systems.

X11 tried to do this with an input-method-controlled overlay, but this 
was incredibly complex and messy and did not really work. I think 
however that the clients can draw this, and often it will be trivial to 
support it.

Assuming there is an input method doing compose-key, here is what I 
would expect to happen if the user types compose,a,e to make æ:

1. User presses & releases compose key.

2. User types 'a'. Input method then sends 'a' to the client. Client 
inserts it into the text as though the user typed it, but remembers it 
came from the input method. This gives the user a preview of what they 
typed, which I have found to be much better than normal "dead key" behavior.

3. User types 'e'. Input method sends a message to replace the previous 
input with 'æ'. The client then deletes the 'a' and inserts the 'æ'.

If at step 3 the user typed 'x' (and assuming ax is not a compose 
sequence) the input method could then choose how to handle the error:

   1. Input method says to also insert an 'x', resulting in "ax" (my 
preference).

   2. Input method says to replace the previous input with 'x'.

   3. Input method says to replace the previous input with nothing 
(matching what X11 compose key does).

Is there anything like this in wayland's input method design?