OpenStreetMap: Is Cartographic Innovation Possible?

The primary way many people experience OSM is through the main website at

Consider what you might have seen, in the map, in 2006. Basically it would be mostly blank with map data here and there where it was available. Incredibly, in the mean time, we’ve built something that looks great today.

But that’s a problem. It looks great.

It’s not great. It’s missing all kinds of address and turn restriction data. Data that is vital to making OSM a true digital map.

It used to be that you saw the worst view of the map, now you see the best view of it. It looks complete. It isn’t!

Here’s what I say we should do: show the worst possible view of the map possible and let others show the best view. When there were big empty spaces in the past people would feel compelled to complete the map. You could see there were blank areas and go fix them.

Now there isn’t that same compulsion. I say bring it back.


By making the map look blank again. Don’t show map data that is fresh. Don’t show roads with no addressing data associated. Don’t show unedited TIGER data.

In reverse order.

TIGER data which has not been edited will have dave_hansen or something as the username. So it’s super easy to filter it. Don’t show that map data at all in our front page rendered map. Or, if you want to, show it in bright angry orange. Call attention to it, or remove it. Create a big incentive for people to edit it. If it doesn’t need editing (and let’s be honest, that’s rare in TIGER data) then we can use the tiger:checked key (or whatever it is) to mark that it is ok.

Address data. If roads have no address data don’t show them. Instantly large chunks of the entire world will go blank. Good! We need a reason for people to collect the data. Or, if you like, show those roads in bright purple. If a road has no addresses on it, mark it as addressing:none. Let the renderer figure out to show roads which have no addressing. Will people add one address point and suddenly the road is visible? Yes! But that’s a good thing. Now up the limit to needing more than one point. And so on. Until it’s marked as ‘complete’.

If data is version 1, that is if it’s been entered by one person, and it’s been sitting on the map for a year then don’t show it. Or show it in bright orange. Force people to go check old v 1.0 data and as above, check that it’s correct in order for it to be rendered.

Instantly you’d have a global map of orange or missing data and thousands of people would plunge in to fixing it all.

Will any of this happen?


Or, at least, I doubt it.

With everyone in charge at OSM nobody is. Innovation, bold innovation, doesn’t happen by committee. Who would give permission for such a bold change in cartography? Who would order it done?

Inherently there would be a gigantic discussion on the mailing lists about the pros and cons by people with nothing better to do. Nobody would feel the authority to make such a striking change, which is (one of the reasons) why we end up with effectively no change in OSM’s user experience. To be clear, I give you permission.

And yes, anyone could go make a map style like this. The point is to make it the default on the main OSM site. You could turn it off if you wanted. There could be a banner saying “Hey, this is a view of our map with all the bugs exposed, here’s how you can help” to manage expectations.

We desperately need those with the keys to the castle to feel both the ability but also the permission to innovate in new and unexpected ways like this to force the project forward.

15 Responses to OpenStreetMap: Is Cartographic Innovation Possible?

  1. right December 13, 2012 at 6:31 am #

    right, and that the reason why they removed redaction bot view from osmi.. sigh

  2. Harry Wood December 13, 2012 at 6:32 am #

    There are a few transitions in progress. Nothing quite so bold it’s true, but…

    A year or so ago the sysadmins formulated a concept of featured tiles. A set of rendering styles (hosted map tiles and tile rendering systems) considered worthy of inclusion as a layer in the top-right selector of the OpenStreetMap front page. We has some layers there for a long time of course, but now the sysadmins have clearly stated criteria, and hopefully it’s clear that the process is open, though this would benefit from being more widely promoted perhaps.

    As you’ve said, *anyone* can set up a rendering style, and you might choose to present the view of the worst aspects of the data (i.e. a rendering as a quality assurance tool). We currently don’t have any such layer featured as an option. This would be welcome. So it’s worth pointing out that *anyone* can also have their style included as an option on the front page. There’s massive hurdles to this still. You need a beefy server to run a worldwide synced-up rendering system, but at least we can also say that…

    …The software for doing this is getting easier to install and configure. Tilemill is bringing this in the right direction. Every time this gets easier, it gets us closer to where we want to be: Supplying data for many many different rendering styles around the internet, with less pressure on the ‘standard’ front page style. But speaking of which…

    ….A very significant development specifically aimed at driving innovation of the ‘standard’ OpenStreetMaps style, is Andy’s work to port the ‘standard’ OpenStreetMap style into carto for use within tilemill. (Taking place very recently)

    The old name of default map style used to be “Mapnik”. This was renamed to “Standard”. This is a subtle change, but again it is aimed at getting people to think about the idea of there being many different styles rendered by mapnik or other software.

    There’s a little way to go still, but “those with the keys to the castle” are accepting patches and, in the ways I’ve mentioned, developers are gradually working on making it easier and more welcoming for people to propose new styles and changes to the “standard” style.

  3. Alex-7 December 13, 2012 at 7:12 am #

    I was collecting today the house numbers in this area– with “OSMpad” for “iPhone”.

    It is the great application, which automates collecting addresses.

    I see future innovation in such applications. One cartographer can map several streets per one expedition. The limit is the battery life of “iPhone”. It lasts about 2 hours in the field.

    What I am missing in “OSMpad” is a possibility to record building:levels. If I am on the ground, near buildings, why collect only house numbers? And not building:levels at the same time.

    I solve it so far like this. For example, house number 15, and levels 6. I type in 15/6, but it is not a clean method. I would prefer one more field in “OSMpad”, namely: “building:levels”.

    So that one does not have to think or to invent something on the fly, but collect data solidly.

  4. right December 13, 2012 at 7:32 am #

  5. AJ December 13, 2012 at 12:11 pm #

    > If roads have no address data don’t show them.

    I kind of like this one.

    Another couple super simple things the main style could do to highlight problems and make certain areas look empty:

    – import with the `–exclude-invalid-polygon` option
    – don’t render deprecated tagging hacks like `natural=land`

  6. karl December 13, 2012 at 8:51 pm #

    The missing information challenge is very interesting and I guess at least worth trying. There will be rough edges to fix along the way, but that could be fixed along the way. For example, most of the streets in Japan do not have name, the urban information unit is blocks of buildings surrounded by streets and super areas. I’m pretty sure there are plenty of walking path without names. Anyway it doesn’t matter, it’s about in each areas what you consider a complete set and showing or not this set.

    I was also thinking about a heatmap of completion. That would also work at many different scales. The more red the area is, the more the map needs editing. The closer you get from details, the better the heatmap becomes.

  7. Andy Allan December 14, 2012 at 3:51 am #

    “To be clear, I give you permission”

    To be clear Steve, nobody needs your permission.

    • Steve Coast December 14, 2012 at 6:24 am #

      I think you took that a bit literally. The point is that progress stalls waiting for some kind of consensus. Also sad that out of everything in the post, that’s what you decided to jump on.

    • Burak December 26, 2012 at 11:37 am #

      Robert,Do you mean that an error-filled OpenStreetMap might be even worse than no map at all? Its certainly true that no map is pefcert, but a than the popular Google maps.Of course, the largest error in OSM at the moment is still the blank spaces that have not yet been mapped. The point is, however, that anyone can fix things in OSM what can you do when Google maps are full of errors?

    • Russ Nelson January 16, 2013 at 5:45 am #

      I thought it was perfectly clear that Steve knows no one needs his permission.His point was that nobody should be waiting for him to give his blessing to a project.

  8. Peter Miller December 14, 2012 at 7:29 am #

    We have a bunch of map views to encourage the collection of map data here.

    The maps are grouped into areas if interest, for example: ‘Highways’, ‘Railways’, ‘Sports and Leisure’ . For most of these there are maps to highlight attributes that are missing.

    This map for example emphasises railway lines that are missing an electrification tag and this shows TIGER data that has not been reviewed:

    Here is a wiki page about the product

    We also have a web service which we have been running for the past 3 years showing where road names are missing from OSM in GB based on open OS data. We are currently at 92% completeness with a distinct north/south divide in England. It shows that in the past month 11,000 additional road names have been added, with 60,000 left to go.

    Feel free to have a look around (no sign-up needed) and do a post about it 🙂


  9. Tom Chance December 14, 2012 at 8:30 am #

    Steve, I think you’re 100% right about the importance of the standard map, but your proposed solution won’t help much.

    Look at the cycle map, it’s right there as an option on the OSM homepage, is used by CycleStreets and has been known about for years. But we still have really patchy coverage of things like route relations. Just showing what’s missing doesn’t make everyone go out and map stuff. Addressing, for example, is really time consuming to map compared to the fun of being able to map a whole neighbourhood’s roads and basic amenities in a weekend.

    Also, the brilliant OSM Inspector and ITO Map tools show that there is far more data to be added than could be shown in one map, or a handful of maps.

    Personally I’d like to see an OSMF-led redesign that gave prominence to really top quality tools like OSM Inspector and ITO Map, perhaps rotating a project of the month (weekly seems to regular) like Map Roulette or “look at this ITO render and fill in the speed limits”.

    As for the reason for the intransigence you mention, surely it’s the result of your mantra of flat organisation and “Just Fucking Do It”? If we had a group elected by OSMF members who had a mandate to take charge of these issues, they would feel empowered to do so. Instead, brave people make tiny changes or suggest big ones and get swamped by the anarchic everyone-has-a-say do-ocracy.

    • Steve Coast December 14, 2012 at 2:12 pm #

      “but your proposed solution won’t help much” … “Just showing what’s missing doesn’t make everyone go out and map stuff.”

      You have no idea whether these are true statements. I have some anecdotal evidence that it’s not true based on how the map evolved from 2004-2008 or so. What we need to do is test it, but we can’t since there is nobody to make the decision in the current organisation.

      “it’s the result of your mantra of flat organisation”

      Yes, you’re right. The mistake I made was giving up technical ownership. I’m certainly not as technically masterful as many of those involved today but that’s vastly outweighed by an ability to listen to feedback, break with tradition and try new things. For an inverse example, you could use swapping out one JavaScript slippy map library for another. While nice, probably prettier code and it might load faster, it has approximately zero to do with end user satisfaction like, I don’t know, a search box that worked or the ability to route from a to b.

      • Harald December 17, 2012 at 11:16 am #

        >While nice, probably prettier code and it might load faster, it has approximately zero to do with end user satisfaction like, I don’t know, a search box that worked or the ability to route from a to b.

        I don’t necessarily agree with everything in your post, but this is so true! I’m a regular mapper and love OSM. But if I quickly want to look up an address or figure out how to get somewhere I virtually always end up with the Google. And that’s sad.

  10. Russ Nelson May 26, 2013 at 7:53 am #

    tag is Tiger: reviewed, and if its present JOSM highlights the way with a big fat ugly yellow background. But yes, mapnik should do the same thing.

Powered by WordPress. Designed by WooThemes