Re: mboned: UDP port conflict mtrace/traceroute

Hi, all, 

On 2019-07-25 15:05, Toerless Eckert wrote:

> I think it would be great if we can capture in an mtrace errata for
> later memory the reason why practically the method of per-packet port
> number scheme in traceroute is highly useful and also widely
> deployed.
> 
> The reason is that you want to parallelize per-hop packets so they
> can overlap in time. To do this, you need a unique identifier in each
> packet that can be recognized in the ICMP unreachable reply going
> back to the sender. As ICMP APIs are not standardized, those
> APIs may not always deliver all fields of the actual ICMP unreach
> packet,

Understood up to here... 

> so the destination port number is the safest identifier to
> rely on.

Why wouldn't the IP ID be equally "safe"? (you can set this even for
IPv6 that aren't fragments). You're supposed to get the Internet header
back... 

> Ultimately, we should ask for IANA to put for the 254 port numbers
> following the traceroute assigned port number something like
> "reserved for tracroute", and maybe someone who wants an easy RFC
> number could write a one-pager outlining this scheme informationally ;-)]

Right now, traceroute already has one and so does mtrace. That ought to
be enough, per RFCs 6335 and 7605. 

Joe

_______________________________________________
MBONED mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/mboned