Realized I left out an important part:
I frame - contains all information needed to decode the frame
P frame - is a predicted frame based on the I frame before it, some information is encoded
B frame - is a bi-directionally computed frame based on the frames before and frames after the B frame
GOP - group of pictures
"normal" GOP of 15 for IBP encoding (1/2 second between I frames)
IBBPBBPBBPBBPBBI.........
You can also get IP encoding which is a larger file, and as mentioned all I frame encoding which is much larger
DV25 is 25Mbps compresed video, all frames encoded
DV50 is 50Mbps compressed video, all frames encoded
MPEG25 is (generally) all I frames at 25Mbps
Mpeg looks at the video to see what has changed and it puts those changes in the P frames, then computes what would need to be changed between the nearest forward encoded frame, and the nearest backward encoded frame, these are the B frames. This makes it hard to edit on a P or B frame since those frames do not have everything needed to decode the frame. So the editor must decode that GOP to temp space in order to edit on those frames. You'll find that most DVD authoring software will only allow chapters to be placed on I frames, and applications that work together like Final Cut and DVD Studio Pro will cause I frames to be encoder when you put the chapter markers in the Final Cut timeline and then export the project to DVD Studio Pro.