On 2/13/12 4:22 , Fred Baker wrote:
Question for you:
RFC 4884 says that ICMP/ICMPv6 define an Echo Request/Reply message, but as I read it doesn't permit the use of that structure with the Echo Request or Reply. It seems like RFC 4884 would help in this context.
For example, if Atlas Probes and their servers use NTP to synchronize time, an Echo Request could contain the time the message was sent, and the service could deduce one-way delay. The response could similarly include a timestamp, the probe could calculate one-way delay, and report the time and delta-time of that response in its next request. Additionally, the service could command a probe to additionally traceroute to another probe, and the probe could later report its experience.
What would it take to facilitate the use of RFC 4884 extensions with ICMP/ICMPv6?
At the moment, the Atlas system does not include targets with special software. We rely only on well defined protocols that are up and running on the Internet. Note that for time measurements, we could just use the ntp packet format. At the moment, the probes also do not run ntpd. The current time synchronization is in the order of seconds. So for one-way delay measurements, the first step would be to enable ntp on the probes and see if on the whole the synchronization distance is low enough that these kinds of measurements would make sense.