The headline is really unfair. The US military creates, at tremendous expense, enormous cartographic knowledge, and makes great chunks of it available to everyone for free. Then entirely separate people use those free data sloppily, and some stuff happens.
This seems relatively easy to solve, technically. The more IP addresses at a given location, the blurrier its location should be. If a thousand addresses are recorded as coming from the same place, then either it's a glitch or a corporate office. Either way, just going there and harassing the first person you see won't help. If a million addresses are recorded as coming from the same place, it physically can't be smaller than a city block and is probably bigger than some towns. Searching for these IP addresses in that database should just turn up a region or state and "no more detail available." Yes, I know that virtual IPs are a thing, and that anyone with a picture of an area can just mathematically find the center, but if you know either of those things the reasons they aren't meaningful seem more obvious.
IANAEngineer; if a solution to a problem this big seems obvious then there's probably an obvious reason why it won't work. But it seems like the people in charge of these databases could just decide not to provide that level of detail for very common IP addresses, and if they did it would have no effect on legitimate users and would be very helpful to John, Ann, and all the private investigators in their area.
It seems like everyone now expects a latitude and longitude, and this company took the path of least resistance and abandoned original plans to stop providing those, just saying "please don't leave out the error radius when you pass on our info!" (and also changed the actual coordinates to stop pointing to people's homes).
Maybe instead they could provide a spatial polygon, literally drawing the error area on the map.
2: according to the article, the people providing the data actually did this: when they had a vague location like "Pretoria" the database entry was "this bloke's house, plus or minus 20 km in any direction". So the data providers actually did pretty much exactly what you suggest. The problem came when people bought the database to use it in other apps, and stripped out the "margin of error" information from the results provided.
Why not, instead of "this bloke's house," just pick the nearest Starbucks when the data show a residential location.
Vaguely on topic, I loved this thing from when the US was doing its first global geodetic survey thing in the 1960s: the abstracted scientific perfection versus where the ground stations actually were.
Ideally, they should use the datum point for each city. But that could be really time consuming. Also, roughly-the-middle is more useful for a lot of purposes, and a lot of places probably don't have datum points.
This is so totally my wheelhouse -- though I deal more with older AMS (Army Map Service) paper maps and GSGS (the British equivalent) maps and the like. But it's the beginning of the weekend here and now that they've doubled the price of booze here it's one of my two days out of the week to drink and I have been drinking so it will have to wait...
How do they double the price of booze so quickly?
It's a monarchy, they do what they want.
Oh, and Herbert Kitchener, the guy who possibly caused more huge guns to be pointed at more South Africans eating dinner than anyone else, started out making survey maps.
Practice for pointing guns at Germans.
Practice for navigating around mines.
They don't take up that much space. The boxes aren't real.
He should have established a long series of complaints with the local police, getting to the point where they realized anything at his address was a false positive, and then sold the house to kidnappers.