Hacker News Reader

Comment by freyfogle | Hacker News Reader

We had a great example of this in a recent Geomob London talk: ask Google Maps to route from Wairiki, Fiji (-16.8048, 179.9893) to Welagi, Fiji (-16.7349, -179.9439), two villages on the coastal road of the same island and about 10km apart, but as you can see from the coordinates on opposite sides of the anti-meridian.

You get "Sorry, we could not calculate driving directions from "Wairiki, Fiji" to "Welagi, Fiji"

https://www.google.com/maps/dir/Wairiki,+Fiji/Welagi,+Fiji/@...

View on HN

It astounds me that this is still a problem in 2023.

I company I coded for surveyed the entire Fiji region in 1997, drafting aircraft at 80m ground clearance in long lines across both sides of the 180 longitude line.

Every single bit of commercial US GIS software available at the time was garbage and failed to correctly compute line lengths that straddled the 180 longitude line, areas of rectangles that did the same, correctly polygon mask etc.

Fortuneately we had in house software (that had the same flaws, considering it originated from Sunnydale, California) and source code so this became Yet Another Patch I had to write in on the fly in the midst of everything else to move forward.

I wrote issue tickets on this to ArcView | ESRI (of the day), the open GIS groups of the day, and raised it with anybody on relevant IRC boards | early reddit who was talking about their aircraft control software, etc.

I haven't thought about that for 14 odd years now .. but I really thought the base libraries in Google Maps would correctly handle geodesic geometry by now.

I'll bet if SV wasn't such a hotspot that such things would be handled better. You can't dogfood this if you're in Mountain View and chances are that the intersection of 'people working on Google Maps' and 'people that use Google Maps in Fiji' is empty.

That would make sense if Google Maps wasn't 18yrs old, but I don't really have opinions on that sort of thing as I've never worked for a large tech company on a mature product. Maybe there's not much motivation to try that hard after a certain time.

Yes, I think I should have made my point better: if you can't dogfood it and the population affected is small people just don't care, especially if there is zero economic incentive. It sucks that it is like that.

Interestingly they're not far (relative to myself in Australia) from Scripps in San Diego .. the ocean research company famous for dolphin research and not ocean tempreture mapping to acoustically track submarines in the cold war.

Scripps authored some pretty decent open source CLI unix mapping software that you could pipe together to do all manner of things data presentation wise that dates back to the 1980s at least.

That could handle line lengths across the 180 longitude line .. they thought globally and ran sensors from California to Hawaii to the South China Sea (and elsewhere in the world).

Even the original Sydney-based digital mapping start-up Where 2 Technologies that spawned Google Maps (not the 3D Google Earth) could handle those calculations (as I recall, they even pulled some libraries from people I worked with) - I'd have to check but I believe Where 2 shared some DNA with ERMapper (Perth, W.Australia based global resource mapping company).

At some point post aquisition they must have rewritten | simplified assumptions and never availed themselves of US open software paid for by their own tax dollars.

Your software was written in the Buffyverse?

Fair point - the original core version from the mid 80s(?) was written by TerraSense Inc. of Sunnyvale, California.

In the above I was recollecting a code header from ~ 26 years ago, I'm Australian and got a single letter wrong in a district name of country I don't live in.

You okay with that or are you just here to nitpick?

Just nitpick.

Interesting, is the ring road around Tavuki Island THE only road that falls on the anti-meridian, at all? A quick look along the line and it looks like it.

It's wild that on google maps, even dropping a pin on the Tavuki road and trying to drag it across the line, it just stops, hitting this invisible wall.

Something about this experience makes me want to go there....

Wow, even zooming in and out on that island makes Google Maps skip into the middle of the ocean.

Edit: and the nearby Rabi Island has a discontinuity at the meridian in map view. Zooming in on it in satellite view sent the map into a crazy refreshing state with flickering grid lines.

These real-life effects by our decisions in technology sometimes half way around the world are very interesting to me. Similar to how some poeple are unlucky enough to live in their communities geographic center that becomes the default location for GPS lookups for that location.

There was US mapping software (not the international scientific survey semi open source stuff, commercial mapping for US public) in the late 1980s and 1990s that actually had positive longitude displays over the US mainland so that their users didn't have to think in negative numbers.

As I noted elsewhere in this thread, large swathes of US software couldn't handle mapping computations that straddled the date line and it's breathtaking that this is still an issue today.

Weird considering that Google maintains the S2 geometry lib which is supposed to handle things better than naive lat/long

> "Sorry, we could not calculate driving directions from "Wairiki, Fiji" to "Welagi, Fiji"

And how many potential users does this affect? Enough for Google to spend 10 seconds worrying about it?

I can't get any routing on that island at all, even within the same settlement

Reply on news.ycombinator.com