Why has the far page$-$replacement strategy not been implemented in real systems?

It is impossible to determine a particular process$$s reference patterns.

The execution$-$time overhead it requires is too great.

It has been shown mathematically to perform at far below optimal levels.

all of the above