Hacker News Reader

Comment by lucb1e | Hacker News Reader

The article doesn't say, but if it used an OpenStreetMap-based routing engine (which you can self-host because, y'know, open data) then the directions might misalign to the display layer which is Apple Maps. Therefore it must send your directions request to Apple. Which I guess is fine, they'll just see "some DDG user" requesting certain directions and it's definitely not the worst company they could be working with, I'm just a bit surprised that DDG went with Apple Maps instead of self hosting.

View on HN

It's the one thing that's kept me off DDG - the superior choice of OSM exists. And as you've pointed out, their choice of maps now forces the user into Apple's hands as the choice of navigation, a company which I do not trust, and I certainly don't see it as the least evil.

Even placing my mistrust aside, it's once again a US-centric move; as garbage as Apple Maps is, it's even worse outside the US.

The only reason for this choice that makes any sense, is that they've received a monetary incentive for this, and neither company has been open about it.

What about factoring real time traffic data into route planning?

OSM Data/Routing engines alone don't address that side of the problem.

Speaking from experience, trying to deploy a production-grade OSM instance is an enormous amount of work. Several businesses have been built purely around that (Maptiler and Mapbox, to name two off the top of my head). I agree that, particularly outside the US, OSM may seem theoretically to be a better choice, but I'm more inclined to assume that DDG simply decided the development resources needed for OSM were too great to justify it, and went with the alternative that they think gives them the best shot at both improving user privacy and competing with google.

That sounds logical to me. Do you see anything the OSM community could do to make this easier? Or do we just need (more, perhaps) companies like Mapbox and recommend those rather than self hosting?

For my personal application, I needed access to the actual data, and wanted my map render to be in-sync with the data layer. For that, there aren't any companies (that I'm aware of) offering services; that would also be really tough to do, because your database data model/schema for that is always going to be application-dependent. So I think that Mapbox and co fill that particular role fairly well. But, I don't believe they offer route planning (it looks like there are other companies that do, but for DDG, their SLAs would be important), and route planning needs access to the raw data as well. So really I think the biggest impact would be in improved tooling.

If I had the time, I'd personally focus on writing a loader replacement (ie, imposm/osm2pgsql "competitor") that:

1. supported arbitrary data transforms into postgres -- basically, can load data into any database schema, do on-load data transforms, enrichment, i18n, etc.

2. was distributable via some distributed message queue (I'd personally choose SQS, but you could do something like celery for less-opinionated solution)

3. was not opinionated about deployment compute environment (ie, neither requiring nor forbidding containerization)

Basically, the ability to have a proper, production-grade, not-just-meant-for-slippymaps ETL pipeline on OSM data would be a really big deal. I haven't ever seen an open source solution for that.

Does OSM have traffic?

So much can be gleaned from the way points. Google guesses home and work with great accuracy without the need for crazy fancy data analysis.

But only because they can tie it to a user account. If DDG is the middleduck acting as an anonymizing layer, Apple won't be able to perform such analysis on the data.

This is all conjecture obviously, it's hard to know exactly how much data Apple is getting, and if they could use something else to fingerprint users, but I'll give DDG the benefit of doubt for now.

Correlate navigation requests made through DDG with location data from people with less of a focus on privacy. From there you can use a social graph to narrow down who lives at a given location to a small set of people. It's a lot better than straight up telling a third party which account lives at/uses a certain location, but it's navigation is a very revealing activity no matter what.

Navigation requests are very frequently home<->"place I went". Household-level driving/shopping patterns are incredibly valuable in their own right, but it's also easy for a dedicated company to tie your address to the personal identities of the residents (Apple's internal customer records, public records, private vendors whose whole business is selling this info), and then it's just profit all the way down.

Hypothetically, DDG could guard against Apple having useful navigational histories for your home by seeding requests with misinformation -- artificial navigation requests Apple can't tell apart from your real requests -- but I'd be surprised if this hypothetical is actually happening.

Only one person (or perhaps two) works at X and lives at Y. Location data can never be anonymous.

They could use Valhalla. It is the open source routing engine Tesla uses in vehicle.

https://github.com/valhalla/valhalla

The article states the directions are calculated by the Apple Maps engine.

Where? When I look(ed) for it, all it says is

> like our embedded maps, powered by Apple's MapKit JS framework

The JavaScript framework they use is not necessarily the same as the server-side routing engine. One can overlay data (like the path to drive and turns to take) with any mapping library like google, leaflet, etc.

You're right. But seeing requests that look like "https://duckduckgo.com/mapkit/?https://api.apple-mapkit.com/......" there is little doubt that DDG forwards directions requests to Apple APIs.

Reply on news.ycombinator.com