I was wondering why the IP protocol use fragment offset to define the fragments of an IP datagram instead of using index number (first fragment with index 0, second with index 1, ...). We already know the size of a fragment with the total length field so it look like a duplicate for me to specifie the offset and (maybe) a waste of space. Using index, with 13 bits we could have 8192 fragments, each of 1500 max bit (on Ethernet), and we could have the same max size using only 6 bits (with a size of each fragment of 1500 bits) or with 7 bits (with a size of each fragment of 576 bits) or 10 bits (with a size of each fragment of 68 bits). Am I missing something ? Is it more easy for a router to construct the datagram with the offset ?