This could be the traceroute faq. It could be V1.00 dated 18-mar-1996 You could get it via ftp://ftp.aces.com/pub/software/traceroute/traceroute.faq You could also send a six-pack of a microbrew beer to gavron@aces.com (address available via WHOIS) and ask him to send you a snailmail copy. You could also keep this copy. It's up to you. ... Q: What is an FAQ? A: It is a series of notes designed to explain the best uses of Traceroute. Q: What's wrong with PING? A: Nothing's wrong with PING. Unfortunately, today's networks are a bit faster, more congested, and more dynamically routed than the ARPAnet of yore for which PING was invented. PING's 1 second resolution, it's unfailing ability to use ICMP echo request/echo response which requires ICMP processing on both hosts, and its lack of simultaneous multiple-probe functionality make it a bit less than useful. Q: What's wrong with TRACEROUTE? A: Nothing. Unfortunately, too many people aren't aware of how to best use it. Q: How do you best use it? A: That depends. Specifically it depends on whether you're using it to determine 1. Routing paths 2. Routing asymmetry 3. Path loss or congestion Q: So tell us A: Glad you asked... 1. Routing paths Traceroute is a fantastic tool for discovering RFC-791 [p.12] compliant routers along the way between you and your destination host. While RFC-791 means "IP" and this would conceivably mean "all IP routers," some manufacturers are too stupid to do this. To use traceroute in this fashion, type traceroute destination-host Example: traceroute ns.internic.net The resulting display will show all "hops" from your first router down the line (hop 1) to the final host. If you see two consecutive hops with the same host IP address, rest assured that the first one violated RFC-791. 2. Routing Asymetry At times, a destination's host's path back to the originating host may go over different circuits than the original path. This is called routing asymetry (or asymetrical routing). It's a good thing because you're testing as much network hardware as possible all at once. It's a bad thing for similar reasons. It also plays havoc with NTP's (Network Time Protocol) abilities to judge path time. To find routing asymetry, one needs to do a traceroute from the destination host back to the source host. This is done using an IP option called LSRR (Loose Source and Record Route). As this is an option, not all manufacturers support it. Livingston is one such bad guy. Usage: traceroute specify-lsrr-host origin-host Example: (VMS) traceroute/lsrr=remote-host-somewhere my.host.domain (unix) traceroute -g remote-host-somewhere my.host.domain These generate a route from you to the remote-host-somewhere and back to my.host.domain. 3. Path congestion or loss Ever since they let graduate students invent software, the Internet has gone to hell in a handbasket. (We're talking about Mosaic and the Web here.) This has left lots more people connected to a much less reliable network. To discover these, it's useful to send out lots and lots of packets, time them all, and see how many are received. Traceroute includes this ability. It can send multiple probes... and it can summarize loss and latency characteristics. Usage: traceroute specify-probes destination-host Example: (VMS) traceroute/number=100/delay_stats destination.host (Unix) traceroute -q 100 -Q destination.host Note: SPRINT's backbone and MCI's backbone and several others use Cisco routers. (This is a Good Thing). SPRINT uses experimental code which is perfectly happy to drop a lot of the duplicate-looking ICMP packets that traceroute uses. Thus, you should IGNORE a loss figure unless you see it continue on all subsequent hops. Example. If on hop N you see 10% loss, and hop N+1 you see 11% loss, and hop N+2 you see 9% loss, you can definitely claim that you think there's a 9% real loss at hop N. Example 2. If on hop N you see 10% loss, and hop N+1 you see 5% loss, and hop N+2 you see 0% loss, you can definitely claim that there is no loss. Ehud -- Ehud Gavron (EG76) gavron@Hearts.ACES.COM : Unsolicited Commercial E-Mail will cost $500/message under USC 47 s 5.5.2 : : which can be found online at http://www.law.cornell.edu/uscode/47/ :