In section 6.2.3, it shows two examples of how Dead Keys are expected to be processed. This processing involves the synthetic generation of composition events. However, in light of Bug 22842, it seems strange now that keydown/up pairs are still expected to be processed during even these synthetic composition process for dead keys.

I don't know what the right solution is here, but it might be along the lines of: 1) Specify the exact synthetic event order in which UA's should dispatch the composition events, and suppress normal keydown/up pairs during this process, or 2) drop the synthetic composition events for dead key input (though I know this is nice for symmetry with other IME processing), or 3) some other solution?

IE9-11 at the moment does not generate the synthetic composition events. We treat the dead key press in isolation, but are unable to report the specific "key" value (e.g., "DeadCircumflex") due to a bug in how Windows manages the keyboard state in this scenario. We are working through the bug with Windows, but wanted to raise awareness of the apparent contradiction in the current spec.