• 18 Posts
  • 307 Comments
Joined 5 months ago
cake
Cake day: May 14th, 2026

help-circle

  • Yep, I think I can see the issue. Let me think thru it properly when I’m back home. The way it works right now is that Docker Compose pulls the Python/nginx/BRouter images (and Kiwix if you’re using that), then bootstrap grabs the CORE WA assets as a once-off download.

    The problem is that on a fresh install bootstrap doesn’t recognise pre-seeded CORE assets if you grab them manually. The ZIM is the exception (that one can be be validated and reused locally). So at the moment it assumes a once-off net connection to grab the PMTiles, search DB and routing data, and then you can basically lobotomize it from the internet afterwards. That was because I didn’t want to package the whole thing as one giant glob with a half GB d/l, as that might choke a SBC (it made my Rpi4 do exactly that).







  • DON’T FEED CANDY TO THE ASD KID

    (but yes, probably, now that I think about it)

    God fucking dammit, now you started me thinking about this. Fuck you very much :)

    • “Monday at 8am, this section normally takes 14 minutes.”
    • “Sunday afternoon, it normally takes 7.”

    Record GPS traces for each road, bucket the results by weekday/time, then feed those historical speeds back into routing. Tada - averaged pseudo traffic flow, barring outliers (like traffic accidents etc). That’s what you’re thinking?












  • No. Reliable, universal and contemporaneous traffic data is generally not open source. You’d need to find a source and populate it yourself (aka the Main Roads department example I gave).

    I can’t do that for every place on Earth. I’m not even sure I can do it for myself. It’s genuinely a hard problem and if anyone has a global fix, I’d be curious to see them add it. Not being snide, actually curious.

    Best route choices: possibly, but it depends on what you mean. Geometry is pretty easy. Inferring best route based on traffic conditions…see above.