So, let's say you have an image file that is big-endian encoded on a
little-endian system. You need to know the dimensions. How do you get
those as numbers the CPU can actually use? Do you just read everything a
byte at a time?
The usual method, at least in my experience, is to do a block read of
the header (ideally, directly into a struct) and byte-swap the various
numeric values from file-endian to machine-endian.
For example:
struct foo_header {...};
foo_header h;
fread(fin, &h, sizeof(foo_header), 1);
h.width = byteswap(h.width);
h.height = byteswap(h.height);
// ...etc.
That means your code is dependent on machine endianness; you don't want to byteswap if your system is already big-endian, same as the file.
Better to write:
h.width = bigEndianToHost(h.width);
h.height = bigEndianToHost(h.height);
POSIX got this right with ntohl/htonl etc., even if the abbreviations are a little cryptic.