4 ms·

Is there a good reason not to use big endian?

• Veserv 2y ago
  For starters, all modern machines are little-endian, so you need to flip the bytes. This is made more confusing when you layer another encoding on top since now you need to define how they are layered and when the encodings apply (i.e. Do you encode the value to big-endian then attach the prefix or do you attach the prefix then encode to big-endian? In this case their explanation indicates implicitly that their mental model is the former. Luckily, I think it does not actually matter in this case.) Big-endian also suffers from problems around streaming unknown-length values as the decoding of a big-endian byte pattern is inherently length-dependent, so even if all machines were big-endian you should still prefer little-endian encodings for wire formats. If you try to pre-load the bytes in a little-endian encoding then the bytes at higher memory addresses, which you want to process later, can be trivially masked off. If you try to pre-load the bytes in a big-endian encoding then the bytes at higher memory addresses must be masked off and then the bytes you care about need to be shifted so they get interpreted correctly.
• nayuki 2y ago
  When I wrote a bigint library, it was much easier to store the array of digits in little endian for arithmetic operations like addition and multiplication.