First of all, in this context, synchronization is not necessarily temporal, but, in a sense, being "on the same page."

Secondly, the encrypting/dercypting process is a precisely synchronous process, like CRC calculation, which NEVER has a conditonal branch. The basic encryption process is (clear-data)XOR(key). The decyption process is (cipher)XOR(key). It's got to be the same cipher and key, but it can be at completely different times and rates.

The first bit of the key must correspond with the first bit of the message, or, at least, the same bitof the key sequence that's encountered at that point in the process at the receiving end as was encountered at the transmitting end. That's the extent to which it must be synchronized. Beyond that, the main issue is that the same key be used at both ends for a given message. XOR and un-XOR are the same thing. Hence, if you xor the key with the clear-message, the result is a cipher. If you xor that cipher with the same key sequence, you will produce the clear-message. If you use two xor's located in the at the same relative point in the sequence in both transmitter and receiver, then you can introduce a cipher-feedback element as I previously described, and the same thing will result, in that the XOR's will still reverse the encryption. However, it is a little more sensitive to initialization. Zero xor i, where i is any bit, will always yield i. One xor i will yield NOT(i).

RE