Been searching and found:

"one nice thing about logminer is that it allows to analyze redo/archivelogs from other databases which means you don't have to be on the production database to do the actual analysis. In your case, i would build the dictionary file from 8.0 database move all archive logs that needs to analyzed to a box running 8i and start the analysis. one caveat is that both databases should be of same characterset / db_block_size / OS"

That right? How restrictive is the last caveat - OS?

Both DBs are same Charset and block size, but one is Windows XP (my PC) and one Windows 2003 (The prod server). Is this the problem?