I was also concerned about the wasted overhead. However I guess it's just there for compatibility (since space is cheap) and for common encodings you'll be able to skip reading it with range requests and use your trusted codec to decode the data. Smart move imho.