Thread: Internet Mapping

Internet mapping turned a remote farm into a digital hell | Fusion

I honestly had no idea there were problems like this.

Frightening. And astounding that it took a decade for someone to figure this out. Even the FBI was clueless?

By the way, MaxMind located my IP address about a mile from my home.

Mine is 15-20 minutes away by car in another county.

Originally Posted by Phyrelight

[...] Now if you have a static IP (usually an extra fee) then that IP should come up pretty close to your house when queried.

And why would that be?

I'm going to guess that to be the case only if the company that owns any particular block of addresses is also matching the (reserved/assigned) IP address with a street address; and then making that information available to third parties (as a courtesy, or for a fee, or as a legal requirement).

But even then...

Do they match IP address with billing address? Do they make an effort to geo-locate the switch/router/bridge/whatever that is getting the IP address? There are lots of reasons to be very confident of a location, but the very nature of IP networking means there is almost no reason to be absolutely sure that the traffic associated with an IP address is also associated with a particular or specific geographical location.

All of that barely rises to the level of quibble. As a practical matter, the company can almost certainly provide a location. The company owns (or leases) both ends of the line, and very few customers are going to do anything other than run a geo-local private network on their end.

But I've seen (and used) some exceptions. The only reason to bring up this hardly-even-a-quibble is to highlight the idea of exceptions. The exceptions are the interesting cases.

Originally Posted by Viktor_Olin

Frightening. And astounding that it took a decade for someone to figure this out. Even the FBI was clueless?

Wouldn't surprise me if some of them were. Clearly at least a few people in the FBI are quite cluefull, but from time to time it seems like not all of them are.

I ran into this problem in 2007 while tracking email spam. The first thing we noticed was that lots of spam was originating from servers located in the Atlantic ocean, about 300 miles off the coast of West Africa. After a few moments of "WTF?", it dawned on me that this was a lat/long location of (0.00N, 0.00W). The admins who kept the geo-location data we were querying had obviously used that as backfill for IP addresses they couldn't lock down.

The part of the article that bothered me most was MaxMind's original design. If feels like their original team (a) didn't understand data and (b) were bad engineers. Central to the design of any process for collecting and organizing data is getting answers to this question:

"What do we do with exceptions?"

There will always be issues and edge cases where the real world reveals itself to be a little more messy than your idealized model.

It seems like their answer to the question "What if we don't know or can't find the location?" was, "We'll assign a default location." ... and then as their default they picked an actual, real, place. This is more than just fucking retarded. This is fucking retarded in multiple ways!

... and their answer to "your default location has wasted the time of tens of thousands of people, wasted valuable law enforcement resources, and made the life of the family that actually resides at your default location a living hell." was, "oh, I guess we'll start using a different default location."

At that point in the article, "Wouldn't the correct solution for your database be to assign IP addresses that have missing or otherwise non-standard locations some sort of code or null-data indicator; and then make that part of your API? That way you can avoid both forward and reverse lookup errors, and also help your clients avoid interpretive errors. Or maybe there's some other method that both records and reveals non-standard cases. I'm not a database expert, but it seems obvious that your method fails a very basic test of correctness. If you don't know the location then you don't know the location. Assigning a location to those IP addresses, any location at all, even one in the middle of the ocean, is - at the most basic level - Wrong with a capital-W. Your business is to provide information. Isn't there a better way?"

... is something that I couldn't help noticing the journalist completely failed to ask.


Æ