Regarding ABR/VBR as opposed to CBR: in my testing, the space conserving with VBR vs CBR was barely significant. I chose many different types of song material to get as good spread as possible. Sure, I used pretty aggressive quality settings, but I would rather err on the side of caution than well, that other side. After all, 2.00 MB vs. 1.75 MB per minute... Moreover, VBR decoding is slightly more CPU intensive (albeit in the end by an almost insignificant amount if I'm to be truthful). I maintain it's impossible to hear the difference between 256 CBR (LAME) and an original CD source, and every scientifically carried out blind test I've come across seems to indicate this is so, even on very expensive audiophool gear, so higher frame size than 256 wouldn't be much use IMO.
As I said before, I don't like the uncertainty factor that VBR adds; the more variables you have, the easier a screw-up would be. Hey, that's some really bad grammar, am I turning into George Lucas?
Regarding ABR: the reason for its craptasticalness is the fact that you beforehand must know the dynamic of the music and calculate what bitrate would suffice, and if you're wrong...well, too bad for you, you won't have much bits left in the bits pool towards the end. Or, if your encoder is smart enough, it'll just not allocate enough bitrate for the highbitrate-craving passages throughout the whole song, to keep the ABR rate at check.
Guest: if you've been following the AAC comments lately, you'll see that it has some serious artefact problems. If that's due to the file format or just crap encoders...well, that remains to be seen.
But hey, look, if Apple chose the format, then it must be inferior. :D