## ## $Id: TODO,v 1.1.1.1 2006/07/08 16:31:44 dave Exp $ ## Optimised Trace When a valid response is received and the probe is not the first in the list, swap the current probe with the first probe. This works on the assumption that if probe 1 is blocked at this hop but probe 2 gets through, the situation will probably be the same at the next hop... If the sucessfull probe is inserted into at the start of the probe list, the whole trace will favour probes known to be successfull... Fast trace When doing a multi-probe trace, don't wait for the individual probes to timeout, just send them all immediatly (might need a short delay). Both packet sending and packet matching code will need to be rewritten to support this. Probe Defaults if no type is specified after a probe type flag (e.g. just "-T" is on the command line) use a default type (e.g TCP defaults to port 80, ICMP defaults to echo, etc). GNU getop_long doesn't really allow this to be done without sacrificing how the command line options are procesed. (when an option is flagged as arguments optional, arguments have to be specified as "-T=80", which sucks...) Protocol trace probably doesn't work Need some method of testing this. But it should already be better than that other *trace tool wot I wrote, that had a big /* FIXME! */ comment in the protocol processing code, etrace assumes any response on the protocol that it's attempting to trace is good. It also might be a good idea to correctly process the ICMP Protocol Unreachable error. At the moment the a Protocol ends when etrace receives an ICMP Protocol Unreachable error. (Which gives the correct result, it would just be nice to do it correctly) Add all the insane packet construction flags It's scares people who read the man page that every letter of the alphabet does *something*, and nobody really uses any of them, but what the hey... Fix ARP handling. I've had to add "and not arp" to the PCap filter because arp requests (especially when spoofing traces from another address) were breaking things. This should not break etrace since ARP should not be part of a trace, but the code should be corrected at some point because this is a pretty naff solution... UberFix: Handle the ARP responses from etrace
