Edited for clarity.
Edited for clarity.
Having low latency from the device seems even more important to me.
First you need to find a way to tap/click things, and not do so with anything accidentally. Not easy. Blink will have many false positives, as will time-based solutions. Having things pop up new buttons (my original approach) becomes distracting as you gaze over many items that you don't intend to click, only to have new things happening that move so you look at them, but shouldn't.
Also eye fatigue comes in pretty quickly. There's something physiologically interesting when things that normally are automatic becomes purposeful, stress is induced. For example, dictation with speech recognition will quickly strain your voice and you can actually get RSI in your vocal chords! Eyes are similar when you use them as a primary interaction driver versus a support.