Well, hopefully the library knows what the host-order is; and network order is well defined. So the function "knows" either to flip (as above), or do nothing....Originally Posted by cpjust
Well, hopefully the library knows what the host-order is; and network order is well defined. So the function "knows" either to flip (as above), or do nothing....
How can it possibly know the difference between Big Endian and Little Endian? Just by looking at a number there's no way to know.
Is 0xFE1265CD02 Big Endian or Little Endian? There's no special bit (like a signed number) that says this is Big Endian...
Output:Code:#include <iostream> #include <winsock2.h> using namespace std; int main() { unsigned long num = 1234567890; cout << num << endl; num = htonl( num ); cout << num << endl; num = htonl( num ); cout << num << endl; return 0; }
Code:1234567890 3523384905 1234567890
Last edited by cpjust; 07-16-2008 at 09:34 PM.
It knows the difference, because it knows the machine. If the machine is a little-endian machine, the number is little-endian. If the machine is big-endian, the number is big-endian.
Yes, but on a Little Endian machine, both ntohl() and htonl() swap bytes around. On a Big Endian machine, both ntohl() and htonl() do nothing.
Right -- ntohl and htonl both do the same thing as each other (assuming consistent bytes) -- but they don't necessarily do the same thing as OP's code. Looking back, your point was probably that ntohl and htonl do the same thing regardless, so there's no need to "choose". I'm pretty sure my point was that yes, sometimes you have to type "ntohl" and sometimes you have to type "htonl", but (1) why reinvent the wheel and (2) it's nice if you can tell just by looking why you're flipping something around.
Yeah, having two names is probably just for convenience. It makes it easier to read code. It seems like they could have just had one function, but maybe coming up with a good name would be a problem.
