The following example seems incorrect to me?
In a simple MZ test app written in fasmw and run in a Windows XP console,
with the following value declared and stored in memory:

MemBuffa dd 00000000h,00000001h

cmp dword [MemBuffa+1],01000000h ;is equal, jne does not jump.
jne somewhere


cmp dword [MemBuffa+1],00000001h ;is not equal, jne, jumps
jne somewhere

You declared which can be seen as
Now you are accessing "dword [MemBuffa+1]", so you are accessing four bytes starting from the second byte of MemBuffa (00, 00, 00, 01).

Remember that Intel is little endian so the first cmp is comparing against 00, 00, 00, 01 and the second cmp against 01, 00, 00, 00.

I clarified your problem? Actually I'm not so sure what are you asking.

Yes, I see. I make this stupid mistake, sometimes. I was thinking MemBuffa+1 dword, instead of MemBuffa+4 bytes, which would have been correct.
The example made it appear that the endian byte order was being flipped, which confused me, further.
Thanks for the reply.