I can't find a good explanation. So here is a simple conjecture:

When utf8 is a code and an encoding / decoding algorithm.

Some character could have the code `11100010 10000010 10101100`, and will be represented the same in any computers. Because the algorithm does not build a number but a set of bytes.

That's how it has to be represented.

Now, for reading, they always look byte-by-byte, and find the character.

When the first byte starts with `111` they have to search for 3 bytes, and then use it for a look up / retrieve the string.

1) Least significant byte first (little end first or little-endian)
2) Most significant byte first (big end first or big-endian).

TCP/IP sockets only deal with bytes so it's not an issue for networks.

The problem is between CPUs that use different byte ordering. The standard is called network byte ordering which is big-endian. Unfortunately Intel uses little-endian so those operating systems must adapt to the network byte ordering and reverse the byte order of numeric data received over a network.

Consider decimal 2864434397 or AABBCCDD as 4 hexadecimal digits (8 hex nibbles).

Unfortunately Intel requires that you store those bytes in reverse order. It only makes sense for internal architectural issues related to the CPU design. It means you as a human have to flip the bytes to read it as we would normally read a number.

I hope that's a more succinct explanation without a lot of noise or rambling non sequiturs.