On Vector Tiles

Posted by Zverik on 29 July 2018 in English (English).

This is a very rough translation of parts of the discussion that happened on the Shtosm Telegram channel. I’ve translated it so the upcoming vector tiles BoF section does not miss my point of view, while I’m presenting OSM Streak in another room.

Kate in her keynote mentioned the recent article by Richard about what OpenStreetMap needs:

The answer is vector tiles, he writes, but before that he makes a long intro about OSM not being a map; it’s a community and possibilities. He is right. But if we talk about tiles, you should understand one thing. Developers of OSM Carto style have been discussing switching to vector since April. They approach it as a technical task: to repeat the same thing, but with a different stack. And Richard’s point is that it would be wrong.

Style developers are fixated on quality, on cartography, on target audience. The make a product: tiles, which are featured on thousands of web sites. And this product, being showcased on our website, is overshadowing the real product that we make: the geospatial database. That is bad, since people believe OSM is tiles and mapping style, not data. They expect from the website to show traffic, to feature a ruler and satellite imagery. People feel that our product is the website, the routing feature, the mapping style.

What we need is to push the style to the sidelines, and to employ vector tiles. Though not to repeat what we already have in them, because then we’d get exactly the same: a product that overshadows the data. We should do intentionally raw tiles, which adapt to user’s needs. Roads, buildings, some POIs by default. But with a few clicks you could see bicycle routes: make a photo and go cycling. A map like a clay: very plain, but take it — and the possibilities are endless.

This change would mean a revolution in how we relate to our data, especially from admins and style developers. It would require us to relieve control to the user. We do that with the data, but we are yet to learn to relieve control for other parts of the infrastructure. The time to learn is coming.

We’re discussing the website. I’m trying to convey the (not very fresh) idea that non-mappers do not need OSM, and don’t even need to know about the OSM. They can use other wonderful websites, a magnitude better in features and mapping style than our website. And that have a ruler. For example:

If a person opened the OSM website, they came not for features and not for tiles, but to learn about OpenStreetMap. When we make a user-facing service with the website, then it’s weird that Tom closes issues about adding a ruler or improving directions. OSM website is not one of many useful tools for a user, but first of all, an instrument for mappers, and a showcase of OSM capabilities. And in that it is awful. For the price of being slightly useful for outside users — and in the end, nobody gains much.

So, the solution is to send outside people to outside websites, built with OpenStreetMap data, and starting to wreck our website into a fun place for mappers. As soon as OSM Carto is pushed back (not removed, just moved into secondary layers), and when we have a map builder on the front page, people would suddenly understand that all this times they were slaves to the single correct way of looking at OSM, and turns out they have options!

Improve the map builder tool a bit, so you can not only create, but also export maps — and you’re looking at a style database for hundreds of new styles. Then, add a style library, and we get styles for cyclists, styles with 100% POI displayed, pretty background styles for infographics, and thematic styles for those looking for a tobacco shop or a toilet, and weird coloured styles for modern art admirers. All that simply because OSM would stop building a third-tier google maps service, and will switch to pushing its unique qualities. For being simple and useful we have MapCat, and being that is not our thing. We can do better.

Comment from kocio on 29 July 2018 at 17:06

Map builder is a nice idea, that’s also what I want to have on OSM website eventually. OSM Carto conversion to vector is a step in this direction, but it’s always good to make migration smooth and that’s I think it’s needed. Another huge problem is hardware - we don’t have too much of it - and the team who would be able and willing to create such map builder. Do you have the idea how should we approach it?

I think it’s time to talk about details, not just general visions. I made a special subforum on rendering maps with vector thread in it - I suggest discussing things there:

Comment from Claudius Henrichs on 29 July 2018 at 19:18

That’s why I always loved the way went the way of showcasing OSM as an enabling project to do stuff™ with geo data. I think the tech transition to vector has the potential to unlock similar avenues to show what can be done with open geo data.

Comment from wille on 31 July 2018 at 22:07

Hi Ilya! It was amazing to meet you in Milano!

So, the solution is to send outside people to outside websites, built with OpenStreetMap data, and starting to wreck our website into a fun place for mappers.

How will we do it? We don’t have a web service based on OSM that is well known by the world. Would we put a banner or a link on the homepage recommending those services? Even for a mapper, calculate routes and search and find a place is something important. A small percentage of the mappers know about those websites and I feel it’s frustrating if them can not see the potential of the data they insert on OSM.

For being simple and useful we have MapCat, and being that is not our thing. We can do better.

I really like the idea of a map builder tool, but if the argument of not being a MapCat is valid, we also can not be a map builder tool, because QGIS, uMap and Mapbox Studio has similar features.


Comment from Zverik on 2 August 2018 at 07:54

Hi Wille, I think you misunderstood the idea of a mup builder a bit. I wasn’t writing about a uMap-like functionality to put custom data on top of the map. Neither I was advocating towards a Mapbox Studio-like full-fledged map editor. It was more about a simple sandbox-like experimental, not too complex map style editor, like the one we had with Cloudmade, but even more geared towards showcasing OSM features. It would require a lot of work on the design, but if done right, it will change the outlook of OSM for the best.

Using the website to showcase common features, like geocoding and routing, has helped in the past. Now a lot of people know you can do that with our map. But also in that we are far behind commercial alternatives, with MapCat being one of the best. So by now we have a very bad user-facing site, and instead of going our own way and making the best mapper-facing website, we struggle with making it slightly less bad for general users.

Comment from wille on 2 August 2018 at 12:45

Nice, I like the cloudmade reference!

Comment from kocio on 5 August 2018 at 00:06

I wonder what do you think about Maputnik? It’s a free map editor using Mapbox GL. Do you envision using it directly, using it as a base for much simpler user interface, or maybe you don’t think it’s suitable for this task? I remember that iD was based on Potlatch 2 architecture, and it still took 0,5 mln USD and months of work for a start.

Any other, probably more detailed/technical thoughts?

When thinking about “showcase of OSM capabilities”, I see mainly how rich OSM data are, and that is what I would try to show.

Login to leave a comment