Is that a very difficult problem? I thought they just grab the location from the phone, send it to their service, and their service looks for other bumps in the physical and temporal vicinity. It's a rather clever solution, but not exactly a technology that seems worth acquiring, unless there are some patents or tricky performance optimizations I'm missing.
This is such a classic HN comment, sorry. You start off with a totally fair premise, asking genuinely if this is a difficult problem.
But then, you go on to speculate not just that it's probably not technically difficult, just "clever", and then go even further to speculate that it's probably "not exactly a technology that seems worth acquiring." Are you asking questions in good faith or are you making claims? What you wrote here seems less a genuine question and more a post to downplay bump's work and their acquisition as worthy.
Can anyone speculate? The only thing I'm coming up with is that it might require a geo index for quickly calculating the distances between a huge number of location pairs.
The edge cases are enormous - smartphones aren't always running on the same time, and their internet connection delays might be different (one on WiFi, the other on Edge) so syncing time would be difficult. One might give a location based on GPS, the other just the nearest telephone pole. Or maybe none at all, or the wrong one (it happens), or different poles. If you get only 1 bump every 10 seconds it would be easy, but Bump's market probably like to hang out at social events, making it harder to figure out time-wise which bumps are which.
I can go on, but if you can't imagine it being difficult, then your imagination needs work!
I think the hard part of their patented algorithm would be to use data from the accelerometer - basically the only thing that can be 'trusted' - to match bumps up through the unique bumping signature/characteristics. Actually - since it's patented it should be public record, so we could go look it up.
From what i understand it uses the accelerometer to create a "bump signature" and matches the two devices according to that. The geo-location thing only helps filter that search and i think they introduced that part later on when they got a huge number of users across the globe.
Now what is note-worthy here is that the signatures wouldn't be "identical". They may not even be opposite as i think the idea is to bump your HANDS while holding them, so the hands may distort the vibrations. So the fact that they actually managed it to WORK is astounding.
It is in same realm as synching datas I think. GPS is unreliable for close distances, phone clocks, user connections, the reading of the accelerometer...every little part adds fuzziness, and you have to trim down your data to get something relevant, limit the rate of false positive while allowing some leeway.
A few years ago a bunch of users in an BBS were playing with the app by bumping alone and see who the algorithm would pick as a match in their town. I guess it got better, but at that time you had something like a one in five chance to get a match in downtown areas.
Not really as far as I was aware as it work(ed) on devices without a GPS fix or hardware. Used accelerometer signals, radio signal strength, as many signals as they could get as I recall. I remember hearing they were even emitting an inaudible tone so that devices in vicinity could pick it up.
