[M4IF Technotes] AC prediction, follow document or follow reference software?

[Lefan Zhong]

Hello folks,
In MPEG-4 document 14496-2, subclause 7.4.3.3, adaptive ac coefficient prediction, it says:
If the prediction block (block 'A' or block 'C') is outside of the boundary of VOP or video packet, then all the prediction coefficients of that block are assumed to be zero.
Beware that statement doesn't tell us what to do when the prediction block is 'inter'. Should we use the prediction block coefficients to do prediction or set to zeros when the prediction block is 'inter' type?
In the comformance test bitstream Hit007.m4v, 1st P-VOP, the 2nd macroblock (coordinate (16,0)), the macroblock is an INTRA block, and AC prediction flag is on, and the 1st Y-block needs to do AC prediction from the left block. But left block is an INTER block, the AC coefficients are: 
-1 0 0 -1 0 0 0 0
1 -1 0  0 0 0 0 0
0 -1 0  0 0 0 0 0
0  0 0  0 0 0 0 0
0  0 0  0 0 0 0 0
0  0 0  0 0 0 0 0
0  0 0  0 0 0 0 0
0  0 0  0 0 0 0 0
Should we use
1024
1
0
0
0
0
0
0
to do prediction or use
1024
0
0
0
0
0
0
0
to do prediction?
In the reference softwares (no matter momusys or microsoft, no matter version 1 or version 2), they all use zeros to do prediction in such cases.
For scence of prediction, it is better to use zeros in such cases. But that is conflict to document.
Thank you for your consideration.
Lefan