Today Jari and Brian explained a whole bunch of things to me about networks! It was fantastic. It's amazing to have people with so much experience to ask questions.
At the end they mentioned that I should look up how
traceroute works and
that it's a pretty popular networking-job-interview question. And I'd just
discovered this super cool Python networking library called
Scapy which lets you construct
packets really easily. So I thought I'd implement traceroute using scapy!
I thought it would take a long time, but turns out that (a basic version) is really easy.
So using scapy, you can create IP and UDP packets like this:
Then you can send a packet like this:
So IP packets have a
ttl attribute, which stands for "Time-To-Live". Every
time a machine receives an IP packet, it decreases the
ttl by 1 and passes
it on. Basically this is a super smart way to make sure that packets don't get
into infinite loops.
If a packet's
ttl runs out before it replies, the last machine sends back an
ICMP packet saying "sorry, failed!".
To implement traceroute, we send out a UDP packet with
1,2,3,.... Then we look at the reply packet and see if it's a "Time ran out"
or "That port doesn't exist" error message. In the first case, we keep going,
and in the second case we're done.
Here's the code! It's 16 lines including comments and everything.
The output looks like:
1 hops away: 192.168.1.1 2 hops away: 126.96.36.199 3 hops away: 188.8.131.52 4 hops away: 184.108.40.206 5 hops away: 220.127.116.11 ...
So it turns out traceroute is kind of easy! Apparently the difference between traceroute on Windows and Unix is that Unix generally sends UDP packets and Windows sends ICMP packets.
tcptraceroute which, well, sends TCP packets.