yes: A - B - C - D are used for the four states for an 90° encoder with an AB output, sorry for the unhappy naming.
and yes: a software that tries to use one encoder input for interrupt the other for direction will not work. You always have to work with OldState compared to a sampled NewState and evaluate all possible combinations then it works. The difference to many other approaches is, that the state of the encoder with a fixed sampling rate. I can't see anything that can go wrong except if the encoder runs too fast for the given sampling rate. And I stick to the statement: an oscillation between two adjacent states will never make a cumulative error.

Werner