If you read RFC for some network protocol, you will notice that word octet is used quite a lot. I hear quite a lot that they should write byte instead. However, there is subtle difference here.
Octet is always eight bits. Although one may argue that byte is also eight bits, but not so long ago one byte had seven bits of data (remember ASCII). If you go even before that, there were some machines that used lower or higher number of bits to represent one byte (6-bit, 10-bit...). At that time, byte was smallest accessible data unit - no matter how many bits it had. Only with very popular IBM/360, everyone moved toward eight bit bytes.
When writing specification of something on really low level (like RFCs do) with need to communicate across different generations of equipment, this difference is important.
Since byte (at current time) is always eight bits, whole argument seems purely academic, but tradition and precision is important. That is why RFC will never use byte when octet is needed.