Historically adaptive feedforward noise compensation using the so called LMS adaptation algorithm has been considered as the basic solution for active noise attenuation despite a number of drawbacks. In fact one can reject unknown and time-varying multiple narrow-band noises without requiring information upon the noise using an adaptive feedback active attenuation scheme. The implementation of this type of scheme uses the Internal Model Principle and the Youla-Kučera parametrization as well as improved adaptation algorithms. For attenuation of broad-band noise, while one still uses a feedforward compensation scheme, new adaptation algorithms have been developed which take into account the interaction between the measurement of the incoming noise and the operation of the compensation device (leading to an internal positive feedback loop which may cause instability). The use of the Youla-Kučera parametrization allows to separate the problem of the stability of the internal positive feedback loop from the problem of optimization of the parameters of the feedforward compensator. This allows also to go beyond the limitations of the length of the compensation path. Experimental results are presented in the last part of the paper.