QUOTE
Originally posted by Air_Borne
Which stuff does still have to be improved on MPC?
I don't think anything really needs to be improved on as in the case of there being some sort of pathological weakness of the format. 99.9% of the clips out there, even the extremely difficult ones, usually encode transparently with either mpc -standard or -xtreme. Buschel himself has even said that the psymodel is "basically finished". Of course there may be possible tweaks here and there which could further reduce bitrate while maintaining quality (I think Frank may be looking into a few things), but that isn't related to a deficiency in quality, only a possible increase an efficiency. This is mostly theoretical though as MPC is already very efficient.
As for other things, MPC will be improving in the area of bitstream robustness and error recovery, encoding speed (with the addition of SSE/2 and 3dnow! code), and perhaps a few other areas like the ability to encode multichannel files as well as supporting a wider range of sampling rates, and the addition of a new native tagging system. I believe some of these things should be included in the SV8 MPC bitstream revision which should be out sometime in the coming months.
QUOTE
will the quality improve much in the near future?
I doubt it. There's really not much room left for quality improvements (at least of anything remotely considered significant) since the current encoder already provides such a high rate of transparency to the vast majority of people on nearly all signals.