What is the advantages of constant bitrates an of variable bitrates? I read some where, Twin VQ being a constant rate encoder offers "better error resilience due to transmission errors.. ??" I don't really understand this..

there is no advantage in using CBR except that filesizes are absolutely predictable (not to mention some better compatibility with flawed hardware decoders) ... CBR results in non-constant quality and therefore shouldn't be used anymore since modern codecs are fully optimzied to give you constant quality at VBR ...

Error resilience is not any better with CBR cause error resilience depends on the data structure, not the encoding mode ... see FLAC which is VBR, compared to Monkey's Audio (VBR as well) ... both lossless codecs work similar, but one of them (FLAC) is better when it comes to error handling or resilience.

Well, cbr could have some advantages when transmitted over fixed-bandwidth channels, like isdn. After all, one of the mp3 goals was transmission over 2 isdn channels.

The question in this case would be to know if the quality would be better using cbr or abr, considering that in this case the abr would have to be controlled regarding bitrate.

... which implies the need for constant practical bandwith ... in germany, the TELEKOM as the major ISP was not always able to guarantee practical ISDN bandwith within the theoretical limits ...

Nobody could, because it does not depend solely on provider - however, provider can guarantee you maximum link capacity till the border with other ISPs, but that is not end-user agreement for a ISDN/DSL, but a business agreement with SLA inside that guarantees QoS - and that is couple of times more expensive.

So, I'd live with "probably 768 kb/s DSL" than "guaranteed link 768 kb/s with SLA" which is 10x more expensive


Anyway, back to CBR/ABR/VBR issue - well, the first issue is fixed bandwidth thing.

The second issue is also concerning streaming, and that's "buffering" thing - i.e., if you have fixed bit rate channel, you know that you have to buffer at least number of bits of the bit reservoir.

This is important for any streaming application, because if you are going to stream your movie file of 512 kb/s through 768 kb/s network, and the buffering is too small, it probably won't survive a 2 mb/s local peak in the file You'll see famous "buffering......" message again, and that's not good

The third issue is low-delay communication - if you don't have a link which is unlimited, or considerably higher than the transmitted bit rate - you have to buffer. The smaller the bit reservoir (for pure CBR it would be 0) - the smaller delay that is.

Of course, if you only store files on your HDD/CD/DVD where the possible bandwidth of the transmission channel is ~ 1 Gb/s you can freely use VBR )

How about the issue of multiplexing audio with video? Would constant bitrates makes the job of audio/video multiplexing much easier? (Or Not?)

I think constant bitrate is more efficient in bandwidth utilization especially in broadcast applications. In the case of variable rate, the bandwidth allocated must cater for the maximum bitrate, which can varied significantly..

Well, I am not sure that cbr is more efficient regarding bandwidth use.
It is (at least with mp3) in the current implementations, but I think that abr could be more efficient.
Of course, as Ivan pointed you would need an audio buffer verification, similar to the vbv used in video.