The job of a good codec is to make the file size smaller while preserving as much of the image quality as possible. With motion video there are two kinds of codecs - ones that just look at a single frame and ones that look at multiple frames.
Those that look at multiple frames (interframe codecs) are primarily targeted at distribution as although they can be very efficient, some of the tricks they employ to get that efficiency don't hold up so well to post production, editing and so on. They're predominantly 8bit, and when faced with too-much image detail, noise, or motion, can "break".
Intraframe codecs just look at single frames. They just exploit redundancy within that frame to crunch the file size down.
The major enemy of compression is noise, which being essentially random is essentially un-compressable. When you keep increasing the bit rate, before too long you're getting into compressing this random noise rather than ignoring it, which is why you'll get to the point where you have to vastly increase the bit rate to see a very small quality improvement.
Graeme
