Mercator's world map of 1569 was useful to navigators at the time because it is conformal (preserving angles) even over long distances. You could draw a line between source and destination, read off the constant bearing, and point your ship in that direction for days or weeks, and you'd arrive at your destination. This line is called a "rhumb line", or (and I love this word) a "loxodrome".

Rhumb lines are absolutely not the shortest route between two points. That would be a great circle route, but computing a great circle requires spherical trigonometry, which is something that navigators were loath to mess with (as your bearing changes over the course of traversing a great circle, so it requires constant attention). Great circles have certainly been used by navigators for a long time, particularly in aviation where carrying your fuel is a factor. Now of course, ubiquitous computing and GPS makes great circles easy, and rhumb lines are just a historical curiosity.

But because of the properties of Mercator being coincidentally convenient for online maps, we find ourselves back in the 16th Century with a default projection that is frankly bad at large scales.

#Maps #Cartography

I wonder what Google Earth did — I think that wasn't Mercator?

#Maps #Cartography #OpenStreetMap

I think this logic has to be backwards. We don't use Mercator because the tiles are square. The tiles are square because we are using Mercator.

Streetview has to do spherical manipulations in the browser in realtime, so why couldn't Maps do the same?

KDE Marble has a neat feature where as you zoom out you can see the globe start to curve, and if you go out far enough you can see the globe, but when zoomed into human moving around a city scale it certainly looks like squares.

marble.kde.org/index.php

I haven't seen many other map programs do that and some times it's a useful feature.

Yep - people talk about Mercator like it's the wrong projection. It's not only the right projection, but given the constraints, it's the only projection when you add north-up to what you listed.

We're starting to see globes at low zooms, but that technology didn't exist at the start of online maps. It still doesn't change high zooms, and all conformal maps look the same up close.

I will talk about the Mercator projection in FOSS4G Europe in one week!

I have the impression you are not coming. It's far away from your place. But you can watch the video later.

My beef with OSM is not the projection but that the tile render times can render OSM unusable, whether the maps are stored locally or served. My contention is that the performance hit from so much detail on a tile varies from 'absolutely necessary, the detail is critical!' , to 'I just need the route I should follow and some stubbed turnings, so it works at light speed with NO lags'. Where a user is happy on that continuum depends on what they are doing. A simplified map that prioritizes rendering the essential over task superfluous detail (rendered on a best-can-do basis if useful or just left out if it's distracting clutter - I probably don't need to render footpaths, buildings or business names when I'm following navigation down a motorway (freeway) on a long journey where navigation clarity & responsiveness are much more important than superfluous detail.

I love the premise of OSM but the user experience needs a close look & significant refinement for each task type and persona.