" but doesn't this assume that the computer responds instantly to requests?"

No, the relevant data is not YOUR internal latency, but the latency of all the nodes in the graph outside you, and the response time between packet sent and packet recieved.

I'm just guessing too, BTW.

Build a thick enough map of latencies around you, and the attacker should be able to deduce where you are by pattern matching... Probably at least down to the city block of not address.

The thing is, the latency map coresponding to the graph of the internet would only be viable for very short periods- very much the same complexity problems with weather forecasts. Take it from me, the Internet has 'weather', and the infrastructure is changed-out (binary)bit by (binary)bit, Cable by cable, switch by switch, DNS server by DNS server, etc. Its the proverbial boat that is buit while sailing accross the sea. and I can't imagine that the map is something that can be assembled or updated quickly (thats a lot of mapping- and you can't just read the latency on a node once- you'll need conditional dependencies and averages), or be accurate for any real amount of time given the shifting sands.

So probably to get a fix on someone, you would need to know their IP and general location (city and network provider), then map the latency of their network aggressively with a LOT of $$resources$$, then wait to pick up a hit from their IP (or not even- if you come in with a fresh IP but if the attacker has other invariant data to map you over IPs, like a system profile) to some website (file sharing or 'no-no' site thats been hacked or is a sting).

Do this a LOT of times and do some statistical analyisis on the results.

Then the attacker could get a finer read on the location.

Or you could just subpoena their ISP ;-) Now days, anyone can do that, after the RIAA has beaten them to hell and gone.

DL