Systeme D


What OpenStreetMap can be.

Over the past two years, the biggest buzz among the geo chatterati has been Justin O’Beirne’s meticulously argued (albeit bizarrely formatted) feature-by-feature comparisons of Google and Apple Maps: their design choices, their data, their production processes.

What fascinates me is how the comparison is implicitly phrased. Apple Maps is fighting on Google Maps’ turf, and Justin O’Beirne never questions that. He asks “How far ahead of Apple Maps is Google Maps?”. It implies the same direction of travel. They’re going the same place, but Google is getting there quicker.

Meanwhile, OpenStreetMap goes its own way.

Most OSM commentary focuses on unimportant minutiae (layers, for goodness’ sake, as if it’s still 2004) without seeking to examine what makes OSM unique—and whether that’s still relevant in a rapidly changing market. Could OSM become a dead-end curio while Google, Apple, and an increasingly self-sufficient Mapbox hare off in another, common direction?

OSM’s continuing differentiation from Google/Apple boils down to two points.

First, a non-commercial imperative. Google and Apple (and Mapbox, TomTom, HERE) are beholden to their shareholders and investors. They do what makes them money, which means car navigation. (Once human-controlled, now, increasingly, self-guided. When people ask “How far ahead of Apple is Google Maps?”, what they usually mean is “Who will get to self-driving cars first?”)

OSM, however, isn’t ruled by shareholder value, but by the preoccupations of its contributor base. (We’ll come onto that demographic later.) Whether that’s a good thing depends on what you want from a map. But it’s clearly a point of differentation.

Second, ground truthed local knowledge. Surveying by locals is the gold standard of OSM, building a rich, intricate compilation of contributors’ preoccupations. The painstaking human curation of areas and topics remains unique to OSM.

Neither of these are under threat from Google/Apple. Outsourced quick-fire digitisation of Street View-type imagery in cheap labour countries doesn’t give you this. Nor does image recognition. OSM’s points of differentation remain clear. In OSM’s early days, commentators used the phrase “democratising mapmaking”, and it remains true. You choose what to map; and you choose how to use the map. You participate. Other maps are a one-way street: sure, you can contribute (actively through map corrections, or passively through using a mobile app that phones home), but the provider chooses what you get back.

So OSM becomes a fractal of local maps. The closer you look, the more hidden detail unfolds.

Rethink the map on

We always say OSM is about the data, not the map. But that’s because we already get it.

Any pro-OSM thread on Reddit, Hacker News or elsewhere quickly descends into “but looks pig-ugly” / “but when I type my street address into it’s not found” / “but doesn’t have live traffic”. We know that’s missing the point, that is just a testbed for OpenStreetMap proper, the data that lets you solve these problems. We understand OSM’s direction of travel. Neophytes don’t. They see a single eccentric-looking (albeit lovely), purplish map they can edit. (That’s why everyone’s first edit is adding a footpath with name=Footpath or somesuch, thinking only about how it appears on

Look back to the inflection points for OSM’s growth. Sometimes they come when it gets easier to contribute to OSM: Potlatch 1, Potlatch 2, iD,

But, more often, they come when contributors’ work becomes more immediately visible. The first was when we went from weekly Mapnik updates to near-instant changes. Later, getting OSM on your handheld Garmin, or after that on your smartphone. Increasingly, it’s high-profile sites switching to OSM; both Strava and Pokemon Go resulted in massive bumps in contributor numbers.

The feedback loop is important. Visualisation is important.

Enter vector tiles.

Vector tiles are the ‘fractal of local maps’ made flesh. One dataset, a thousand visualisations. The server sends the rich, raw data; the client slices and dices and presents it any way you choose. Bike map, walking map, skiing map? Kids’ map, salesman’s map, retired globe-trotter’s map? Art deco map, pop art map, antique map? All those; and a thousand you and I have never imagined.

Moving to restylable vector tiles on would explain OSM’s direction of travel more than a year of Reddit and HN comments could ever do. In the words of OSM’s mission statement since 2004, “maps you can use in creative, productive or unexpected ways”.

What might this look like? Think of a ‘style explorer’. Think of a new Design tab. Think of designing a style by starting with a bare-bones template then adding ‘playground’ and ‘park’ and ‘zoo’ and ‘railway’ and ‘dinosaur museum’ (to enumerate my two year old’s current obsessions). Think of storing styles against your user account and forking others’. By exposing OSM’s local, human data, we create a thousand new inflection points for the project to grow.

Let’s explore this a bit more.

Meeting people where they are

People contribute to OpenStreetMap when it’s relevant to them. Outside pure geekdom, cyclists were the first group to embrace OSM. It started with two speculative punts on creating maps relevant to cyclists (OpenCycleMap, of course, and mkgmap which makes Garmin maps), plus a small amount of seed data: I still grin recalling OpenCycleMap’s creator, Andy Allan, commentating (at the second OSM conference) on an animation of OpenCycleMap’s first year “and here you can see where Richard Fairhurst decided to cycle across Wales”. Now, in 2018, every cyclist uses OSM: Strava, Garmin themselves (...and may I mention

The humanitarian cause is the other stand-out. Thousands of people have attended mapathons or slaved over remotely assigned ‘tasks’ in the cause of humanitarian mapping. Most of them will never visit Haiti or Kibera. But they believe in the cause, and they gain intrinsic reward from helping. Again, it’s relevant to them.

You can do all the outreach you want and agonise over conduct and tone. It’s all good, necessary even, and it’s all worthwhile. But if the map isn’t relevant, no-one will stay. The path to diversity in OSM is by building more OpenCycleMaps.

Which is exactly what vector tiles give us.

You know what I’d like as the father of a toddler? A map of playgrounds. Every family would. The answer is already there, hidden in OSM’s data. Maybe one day someone will build a Playground Finder app. But we can shortcut the “maybe” by baking that into

And then families start adding missing playgrounds; and OSM’s contributor base becomes that little bit more diverse. I started by describing OSM’s fractal of local maps as “an intricate compilation of contributors’ preoccupations”. A restylable vector map makes space for more preoccupations, not just those sanctioned by the maintainers of one stylesheet; and by welcoming more preoccupations, we welcome more, diverse contributors.

Eliminating the drudge work

There is no glory in clicking a mouse. OSM’s strength is ground-truthed local knowledge. It’s not magically better because each point has been positioned by hand.

Technology moves on. In 2004 we didn’t have high-res global imagery nor anything but the most rudimentary handheld surveying devices. Now we do, and no-one would seek to go back.

In five years’ time, iD (or its successor) will trace the geometry for you. Mobile apps which recognise shapes from the camera will see a sign and suggest that a speed limit should be tagged at this point. Sensor data or imagery diffing will suggest you might want to go and survey the new road a mile from your house.

This isn’t the boring old “survey vs import” argument. This is about using technology to make contributing local knowledge easier and faster. And, again, by doing that we encourage a more diverse contributor base—broadening beyond people with lots of spare time and computer skills.

All of this circles back to the same vision. A fractal of local maps; a map for every human, not just those who drive cars; a map that anyone can contribute to, and then visualise in the way they choose.

Getting underway

This vision, and this article, have been brewing for a year now. I’ve jotted down disconnected paragraphs in spare moments on trains, in cafes and pubs, occasionally dropping comments on Github issues, forums, Twitter. But now is the time to move forward.

There’s a wide understanding that vector rendering is the next challenge for I’m anxious we should get it right, just as Steve Coast got the concept of OpenStreetMap right, I got the UI of Potlatch (mostly) right, Tom Hughes and Grant Slater have consistently got OSM’s scalability right. This will be another of those decisions which will determine OSM’s direction of travel.

Treating it as simply a back-end change for the existing (lovely, eccentric, purplish) openstreetmap-carto style would be a business-as-usual solution. The maps would be a bit sharper, a bit faster to load. We might manage a few worthy side-wins around internationalised placenames. But it won’t move the needle for OSM. This will.

This weekend (28-30 July 2018) is the State of the Map conference in Milan. I’ll be hosting an informal Birds of a Feather session on moving towards vector tiles at 16.00 on Sunday afternoon in room S.1.6 - come along!

Posted on Friday 27 July 2018. Link.

Previous post: Adventures with the LG 360 Cam.