OpenStreetMap

Pieter Vander Vennet's Diary

Recent diary entries

MapComplete: 2023 in review

Posted by Pieter Vander Vennet on 15 January 2024 in English.

2023 in review

Now that 2023 has come to an end, it is an appropriate time to take a look back and see what has happened within the MapComplete-sphere.

2023 also marked the year that I (pietervdvn) received a grant by NlNet, meaning that I could spend a ton of time on improving MapComplete - and with success.

Looking back, a humongous amount of work happened. I’m giving a quick recap here.

User survey and other statistics

I’ve started the year with orienting myself. I ran a user survey (part 1, 2 and 3) and analyzed how mapcomplete was used. For example, there are some interesting statistics about the number of pictures created and about the reviews that were made

Lots of improvements

Most of the work of course went to programming MapComplete, which underwent a few big changes (notably the UI-framework and Mapping-library) and received numerous small improvements.

A quick recap:

Svelte (Q1)

The first big change of the year was switching to an actual frontend framework. MapComplete was written in a hand-rolled framework, which wasn’t very performant. And while I really loved it, using Svelte made the frontend more approachable for other programmers, more maintainable and faster.

Svelte was chosen partly because it works and has a large ecosystem, but also because it turns out to be conceptually similar to the previous, handrolled framework. Even better: the old framework is so similar, that they can be used together! With a few tweaks and adaptions, they were made compatible.

The big advantage of this compatibility is that it becomes possible to gently migrate. Instead of porting everything at once, component per component can be switched when the time is right. As such, there are still a few components around written in the old framework, but they are slowly getting replaced.

Another important bit here is that I received lots of help, both in making the choice of Svelte and actually switching. I received help from Tobias (and the rest from fixmycity.de), but also from Robin van der Linde and Wouter van der Wal. They had experience with Svelte - which I lacked. They basically did all the nasty configuration files and the adapter classes. Back then, it was as if they used some arcane magic to get it all working ;) Luckily, I know Svelte better know, so I at least know how to do such a tricks. But the configuration files are still scary…

UX overhaul

A designer from FixMyCity Berlin had a look to MapComplete as well; they worked out a conceptual framework to design mapcomplete against. The calmer, cleaner, gray-based design is thanks to them.

Furthermore, they gave some feedback on various flows, which made the app more consistent and easier to use.

MapLibre (June)

Another big change was replacing Leaflet and using MapLibre instead. This was another big change which had some fallout, but is worth it in the long run. Not only does it give more possibilities (such as rotating the map or pitching it for a 3D-view), it also paves the way for more optimalisations later on (which I’ve just begun working on).

Move of domain name (August)

In august, I’ve finally made the switch to a proper domain name: mapcomplete.org. This change had quite some fallout, as MapComplete relies on several external services which suddenly need to be updated.

Studio

Another big, new feature is MapComplete studio, which allows to create new map layers without having to know the JSON-configuration. I have to be honest: it works, but there are still many bugs and inconsistencies. It is not as good as it can be, but I’ll be fixing small issues when someone complains about them (and when I do have time)

Export-functionality

Another neat feature which has been polished up and released is the ‘export’-functionality. This allows everyone to export the visible data as CSV, PNG or PDF, ideal to further process the map data.

There even is a ‘line only’-export function, which, in combination with the cycle-infra theme, is perfect to get an SVG-road map. With some post-processing, this can be fed into a 3D-printer…

Imgur data backup

Imgur - our image host - announced a change to their policy indicating that they might start deleting old, unused pictures. As such, many people were a bit worried and we started backing up all pictures on image which are linked to from OpenStreetMap.

This collection of around 30K images has several copies now with various people.

And, as it turned out, only 6 of them became dead. It seems that Imgur’s policy was mostly applied on a specific kind of pictures, namely the indecent type

New themes and theme improvements

Of course, various improvements to the map themes were made. There are too much improvements to sum them all, but I’d like to give an overview of themes that have been officially added to MapComplete this year:

This is a small amount of themes, especially compared to 2022 when 25 new themes were added. This is partly because I focused on improving the system itself (and not on creating new themes) and partly because there was no Open Summer of Code this year, which traditionally meant the addition of many new themes and layers.

Another reason is that many topical maps already exist by now - especially the broadly applicable ones. As such, there is less need to create new ones.

And maybe I should improve Studio further and officialize a few of those themes. In either case, if you have a thematic map or layer lying around, let me know and we’ll add it to official mapcomplete!

And even though there weren’t many new themes added, quite some improvements were made to the already existing themes. For example, drinking water was improved with more specific questions, one can indicate what services a hackerspace offers, …

Accessibility

One of the last big changes made in 2023 is the effort to make MapComplete more accessible for people using screenreaders, braille devices or similar. This effort is still ongoing (but nearly finished). In a few weeks time, MapComplete went from being utterly unusable to being (nearly) feature-equivalent for blind people.

Community

2023 was the year that the community had a small but healthy growth.

Community calls

In the first 6 months, there were biweekly community calls. They were mostly between Pietervdvn and Tordans, but often joined by Robin or people passing by once to tell about their project.

After the summer holiday, the community calls waned and weren’t restarted - I was to busy and neglected them a bit. But - rejoice: we will restart them! The next one is on friday the 26th of january, 14:00 CET.

Patrons

Another nice milestone is that two more patrons started pledging a small amount for the development of MapComplete.

I’m now receiving €1.03 per week via Liberapay. I’d like to thank my three patrons for the trust and small financial contribution!

If you feel like becoming a patron too, head over to Liberapay.

Mastodon-accounts

Another huge success in 2023 were the mastodon-accounts.

The MapComplete-account had 800 followers in march, which passed the 1.000 mark in August and around 1200 at the time of writing, with a few new followers daily.

Another success is the automated account which posts a small, daily overview of changes made with MapComplete, including up to 12 pictures which were made the day before. 133 people follow this account, and there was some interaction every day (e.g. a star, a boost or a small message)

Another fun fact: the automated account was registered on the 12th of january 2023, so it is a year old now!

Change Metrics

But of course, the most interesting statistics are how MapComplete was used in the past year.

Some basic logging is done with goatcounter, which shows around 200 up to 300 visitors per day on average. That isn’t a neglectable amount, but it is not very big either. On the other hand, I suspect that goat counter misses quite a few visitors. OSM-people tend to be privacy-aware and run scriptblocker, which’ll also block goatcounter.

Also, there is a big gap in December. The script was not properly configured and thus not counting visitors.

However, the most interesting stastic is of course how much changes are made to OpenStreetMap with MapComplete. There is a small, unpolished, irregularly updated statistics page on MapComplete which has some nice statistics, so lets dive in!

Modifications name sorted by theme

Lets start our little analysis by the graph “number of changed objects per theme”. This graph is dominated by two themes. One of them, the ‘GRB’-theme (in red) is quite a special theme which causes thousands of changed objects, on one day even more then 18.000 changed objects by a single person. This might seem like an extra-ordinary amount for MapComplete, which is supposed to focus on individual POIs…

However, the GRB theme is a very special theme which helps to import building data in Flanders. As such, this is a glorified copy-paste from the governmental dataset into OpenStreetMap, which causes this ampount of changes.

ModificatiosPerTheme.png

The etymology theme

The second theme causing thousands of changes is the etymology-theme. Here to, some automation inflates the numbers: when one selects a road and applies an etymology to it, all nearby segments with the same name get this etymology applied, which in some cases can amount to quite some changed segments - often five to ten, but sometimes even 50 or 100 segments.

If a few contributors add data on the same day, this can amount to a lot of changes - one day even close to 10.000

Another interesting fact about the Etymology theme is that it is very popular. Many days had several contributors making changes with the theme; 346 days of the year saw at least one contribution with the change! That makes that no changes with the etymology map were made on only 19 days.

Kalepom, AE35, Mikalet, NaanAvanlllai and Peter Broderson were the five top contributors here.

ModificationsEtymologyPerContributor.png

Modifications name sorted by theme, excluding GRB and etymology

When filtering away those themes, a much more uniform image appears, as a patchwork of various edits by all the maps. Of course, there is some variability - e.g. if a single mapper did a campaign around e.g. benches or AEDs

ModificationsPerThemeClean.png

If we have a look to the number of changesets created per theme, we see a typical distribution where a few themes have a lot of edits, whereas others are less popular.

ThemesChangesetCount.png

The most popular themes are:

  • Cyclofix (988 changesets, red)
  • trees (880 changesets, yellow)
  • 853 changesets without actual changes (these are caused by bugs - should figure out why this happens rather soon…)
  • Benches (701 changesets, green)
  • aed (656 changesets)
  • artwork (630 CS, orange)
  • charging stations (563)
  • street_lighting (437)
  • toilets (416)
  • food (391)
  • advertising (380)
  • waste (377)
  • surveillance (357)
  • shops (347)
  • and then a long tail

What can we learn from these statistics?

First of all, the number of changesets has some indication of actual changes, but it is not directly related. A changeset can be a single answered question or hundred of new items.

Second, cyclofix remains the most popular theme; this is probably because a lot of OSM-people are interested in sustainable mobility, because it is one of the themes I pich as example and because it is has the first position in the index.

Trees follows closely, with 10% less changesets. This is due to the easy to make changes with it - there are many unmapped trees, but also because a few people are dedicated to mapping all the trees in their environment.

I suspect that a similar reasoning applies for the third theme, namely ‘benches’. Benches also have a strong usecase, as it is important for less mobile people.

The fourth theme, AEDs, got a whopping 656 changesets. This might be because the emergency services of the Province of East-Flanders (Belgium) did a campaign where they imported all the AEDs that they knew of with MapComplete. Additionally, Toni Sierra and PhillipeC seem to be very intersted with 84 and 78 changeset respectively.

Some other surprises are in the big number of changesets by the surveillance- and advertisement-maps, which both have very strong political connotations. Especially the surveillance camera map is very delicate, yet 89 people made changes with it.

Who did contribute?

With the number of contributors, we see another, very typical distributions. A few very active people and a long tail of contributors making a few changes: (The graph and statistics below are without Etymology and GRB)

ContributorsChangesetcount.png

The top contributor - by number of changesets - were:

  • 5R-MFT with 1.171 changesets
  • Pietervdvn (me) with 733 changesets
  • vjyblauw with 605 changesets
  • Awo (495 changesets)
  • ThibaultMol (485 changesets)
  • Kalepom (430 changesets)
  • Toni Sierra (362 changesets)
  • paunofu (312 changesets)
  • user_5589 (312 changesets)
  • APNeunzehn74 (277 changesets)
  • Prem Das (245 changesets)

No big surprises here, except that I have been completely overtaken in number of changesets :gasp: Congratulations 5R-MFT! This is the first year ever that I’m not the one who made the most changesets with MapComplete. This is still the case including etymology and GRB themes with 1.174 against 790 changesets.

What languages were used when making changesets?

To finish the metrics: a small statistic about the languages that people use MapComplete with:

Languages.png

The top languages are:

  • English + English-US (4544)
  • Dutch (2360 changesets)
  • German (2.255 changesets)
  • French (1314 changesets)
  • Spanish (837 changesets)
  • Catalan (803 changesets)

No big surprises here for me. As I’m Dutch-speaking and promoted MapComplete in Flanders (and had a few Flanders-specific projects), it is not a surprise that Dutch is over-represented.

Catalan is also slightly overrepresented, but that is because a proud catalan took the time to translate it to catalan early on and uses it frequently.

If you want to see MapComplete in your language, head over to Weblate, create an account and start translating! There is, in settings, also an option to enable ‘translation mode’ in MapComplete. Many pieces of text in the interface will gain a small ‘translate’-button which you can click to immediately translate or correct that piece of text. If you enable translate mode, you’ll also have a button to quickly translate all strings of the specific theme.

Translations are pulled in regularly (about once week) so you’ll see your translations quite fast. As of january 2024, the translation completeness for English and German is (near) 100%, where as translations for Polish, Dutch, Czech and Catalan are quite complete as well.

2024-01-15 TranslationCompleteness.svg

Wrapping up 2023

In summary, MapComplete has become a lot more mature, usable and faster within 2023. The absolute number of changes made in 2023 went up in a sustainable manner - increasing by a factor 1.5.

So, all in all, not a bad result; but it would be nice to see a doubling in number of changes next year! Around 14.000 changesets (!) were made with MapComplete last year, so lets try to get 30.000 changesets together next year!

Sources

Want to see the number for yourself? Have a look at https://mapcomplete.org/statistics

Generating the creation date of themes was done by running a small script in ./assets/themes: for f in $( ls */*.json | grep -v "license_info.json" ) do echo -en "$f " git log --follow --format=%ad --date iso $f | tail -1 done

This is part two which highlights the results of the OSM user survey. Read part 1 about demographics and identity and part 2 about the favourite maps

How well-known is MapComplete?

Not that well-known, it seems. In the previous question, 11 people out of 59 who took the time to fill out this question, mistook MapComplete for StreetComplete. This is a clear sign that there is still some work to do.

How did people get to know MapComplete?

How did people get to know MapComplete in the first place?

Via Reddit (13 mentions), Twitter and Mastodon (13 mentions) and the Weekly OSM (9 mentions).

There are honorouble mentions for online chatrooms (6 mentions), word of mouth (6 mentions), the OSM-forum (3 mentions) or ‘arriving via a specific map’ (3 mentions).

From these results, it’s clear that the online spaces where I regularly pitch MapComplete (namely Reddit and Mastodon) also resulted in some people discovering MapComplete.

However, this makes me wonder how applications such as StreetComplete and EveryDoor got to such a big userbase quickly. It seems that creating a mobile phone app with offline capabilities helps with this.

Good questions to ask next year?

I’m planning on doing a similar survey next year (or in one year and a half) to see how things evolve. To be able to compare results, it is interesting to have the same questions, even though some improvements can probably be made (e.g. in wording and more nuanced options).

It is also hard to gauge if people are part of a marginalized group. As such, it is hard to know if we reach those people as well.

But there might be room for other good questions. If you have suggestions, feel free to let them know

Anything else you’d like to say?

This was the question with the most uplifting answers, as many, many people wrote in a compliment about how much they like MapComplete and the work I did! (Well, some of them were probably thinking about StreetComplete)

Thank you everyone involved!

Conclusions

To wrap it up:

  • OpenStreetMap is a very male-dominated community
  • People use MapComplete because it fits their activities and hobbies, it helps them professionally, as activists, to help other people or for daily issues
  • There are many different needs and wants in the community
  • MapComplete is not well-known yet, we should consider different branding and communicate more
  • The community would like a custom theme builder and offline capabilities (e.g. as an app)

That’s it! I definitively learned much thanks to this survey, hopefully you did learn something as well.

And as usual, feel free to follow my mastodon or the MapComplete Mastodon for more updates.

This is part two which highlights the results of the OSM user survey. Read part 1 about demographics and identity here

Which thematic maps do people use?

56 people gave insight in their favourite maps, yielding a total of 86 mentions of specific map themes or groups of themes.

The theme with the most mentions -namely 9- was the etymology theme. This is not a big surprise, as there has been a tremendous amount of changes made with Open Etymology-map. Some people mention curiosity for their local environment, others are interested in the link between OpenStreetMap and Wikipedia. It should be noted that nearly everyone who mentioned etymology indicates that they have hundreds of edits.

The second place goes to the waste theme with 8 mentions. I have to be honest, this came as a total surprise to me! At first sight, it is a bit of a boring topic - especially when contrasted with something like etymology. Yet, everyone needs to get rid of some waste every now and then. By the way, this theme was created by @Rlin and I have to admit that this is one of the most polished themes on MapComplete, using a lot of the available functionality. Thank you!

The third places goes to all cycling-related themes (7 mentions of ‘cycling’). Cyclofix had another 5 mentions, resulting in 12 total mentions. This was no surprise either, as cyclofix has been historic driver of many changes made with MapComplete. It is one of the oldest maps on MapComplete, yet it is still popular and is embedded on a few websites.

The third place goes to benches with 7 mentions - another very down-to-earth topic but with lots of value and lots of unmapped features.

Out of the other answers, it is clear that there are various reasons to use MapComplete:

  • For practical, day-to-day reasons, such as searching for information about POI (e.g. waste, benches, restaurants, pubs, toilets), commuting (cycling, charging stations) or a professional interest
  • Because it helps them with their hobby, especially outdoor-hobbies such as recreational cycling, climbing
  • Out of personal curiosity or wonder in the world (etymology and artwork)
  • They add data because they want to help other people (benches, AED, hydrants, drinking water, toilets, wheelchair, osm-for-the-blind, adding pictures) - which sometimes even touches upon activism (one person mentioned that Speed limits map helped me to discuss about mobility with my municipality.)

A full table of total mentions can be found below

Map Total
etymology 9
waste 8
Anything cycling-related 7
benches 7
cyclofix 5
wheelchair and blind_osm 5
charging_stations 4
aed 3
trees 3
hydrants 3
drinking_water 3
artwork 3
notes 2
shops 2
restaurants 2
healthcare 2
bookcases 2
toilets 2
ghost_bike 2
street_lamps 2
personal 1
pubs 1
nature 1
Fritures 1
climbing 1
pinjepunt 1
speed 1
surveillance 1
pets 1

What do people still want?

I did ask several open-ended question to gauge interests and for bugs, namely:

  • If you could wish for a map-based application, what would this application do?
  • What features would you like to see in MapComplete?

Question: If you could wish for a map-based application, what would this application do?

This open-ended question came quite early in the survey - even before mentioning MapComplete. This is an intentionally broad question, in order to generate as much ideas as possible and to get a feel of what people’s general grievances are.

56 people gave a very extensive answer to this question, but as answers touch upon different topics, it’s very hard to categorize them.

However, there were a few recurrent themes:

  • OpenStreetMap got often compared to Google Maps. Some liked the convenience of that mapping platforms, where others openly wished for open/libre/privacy-respecting alternatives for Google Maps
  • Search was another common grievance
  • Routing was mentioned often as well. Quite a few mappers want a routing engine, taking them to places that need to be updated in OSM or Mapillary; others have specific needs (such as car-free, indoor routing, with live data or offline)
  • At last, some people would love to see features from some app to appear in some other app as well.

Question: What features would you like to see in MapComplete?

Another open-ended question where people can report bugs and feature requests.

The most popular request is to be able to build MapComplete-map easily, with 7 such requests in 59 answers. Good news: part of the NL-Net grant is building a new Custom-theme-builder!

Furthermore, an offline mode was a popular request with 5 mentions. Similarly, 3 people asked for an installable app. Good news for them too: MapComplete is a Progressive Web App, which means you can ‘install’ it on your mobile phone via the menu. It’s a glorified bookmark on your homescreen - but it opens full-screen. Furthermore, aggressive caching makes that it should still open and show some data for a previously visited location and map theme!

There were 7 requests for various improvements to specific themes and a few requests for advanced features which might or might not fit the concept of an ‘easy to use’-map editor.

Houston, we have problem…

Another problem surfaced in the previous question. A big chunk of answers made it clear that there is still a lot of confusion with the android app StreetComplete - namely 11 answers (or 6% of all respondents)…

This implies that the results of this survey should be interpreted as being about the OpenStreetMap-community, not about users of MapComplete specifically.

And it also brings us to our next topic about how well-known MapComplete is, which you can read more about in the next post!

Follow my mastodon or the MapComplete Mastodon to be the first to know about other news.

The MapComplete user survey results: Part 1: user demography

Posted by Pieter Vander Vennet on 8 March 2023 in English. Last updated on 11 March 2023.

MapComplete user census

As you noticed, MapComplete ran a user survey during january. What did it tell us?

The use survey had a few goals, namely:

  • discovering what demographies are using MapComplete (and which are missing)
  • and discovering what needs and wants are still there

Basic demography

The first question is of course: who did fill out the survey? If we look to the numbers, a clear pattern emerges.

The age distribution looks pretty normal - there is a clear peak around the bucket 30-40, which falls down left and right.

The gender is not as balanced. Unsurprisingly, the majority of respondents is male:

Around 10% of respondents identified as female. This indicates that women are vastly underrepresented in this survey - ideally they would be around 50%.

11 people identified as having a non-conventional gender. Note that there was no option for trans people - they could choose between either their chosen gender or use the ‘something else’. As such, I don’t know how much trans people are in each category. One person, identifying as female however stated to be trans (and I suspect that there are a few more).

Non conventional genders are thus vastly overrepresented in this survey. According to Wikipedia, about 0.5% of adult populations identify as non-binary. With 11 out of 177 respondents using a non-conventional gender, we end up at 6% - or ten times what we would expect if the OpenStreetMap-population would reflect the broader population. But, this also indicates that we are a welcoming community in that respect - or at least I hope so.

Another clear bias are the computer skills. 55% of all respondents indicated that they are programmers - another huge overrepresentation:

Now, this is a bit painful. MapComplete aims to be an easy-to-use tool for non-technical users. The survey clearly failed to reach these people (and maybe MapComplete has a hard time reaching those people too).

This is also mirrored in the question on how good they know OpenStreetMap. Close to 75% indicate to have at least hundreds of edits.

For someone to fill out the survey, they first need to hear about it and find the link to fill it out. The communication and communication channels about the survey are thus important. As the survey has been promoted via Mastodon, this probably had a major influence: Mastodon has a userbase which is both very developer-oriented but also quite queer as it has (relatively) many transgender and genderqueer people. As the post about the survey gained a lot of traction there, I suspect many found the survey via that channel (thus partly explaining those biases).

A second important effect is the language. The invitation for the survey and the survey itself where all in english. Developers are generally fluent in English, but a non-technical user might not bother with a survey that is not in their native tongue. As MapComplete is mostly popular in Belgium, Dutch- and French-speaking people might not feel compelled.

At last, some people from minorities are less likely to fill out surveys (source). I tried to counter this by explicitly inviting those groups to fill out the survey in the request, but this psychological effect is very hard to measure.

As such, while I do think that the data is mostly representative of MapComplete-users and the OpenStreetMap-community, I think that less-technical people are a bit underrepresented.

Identity

A last question in the ‘demography’-set was “how would you describe yourself?” - a notoriously hard question which only 66 persons (37%) answered.

This question is intentionally open-ended, as people will normally state what they find important in life.

18 of them mentioned to be a map lover or OSM lover, 15 self-identified as being a ‘techie’, ‘developer’, ‘engineer’ or similar. 8 found Open Source-software important; 7 mentioned to be interested in environmentalism, urbanism, transportation and/or political issues. Other notable mentions were to be involved (professionally) with GIS, to be a teacher (2), a cyclists (4) or a rock climbers (2). One person indicated that they were ‘disabled’.

about the next post

So, this wraps up the first part of who filled out the user survey!

Read more about what themes people preferred in the next article about the user survey

Subscribe to my mastodon or the MapComplete Mastodon to get updates about further news.

About a year ago, we launched a mapping campaign at the request from Visit Flanders (Toerisme Vlaanderen). This mapping campaign is focussed on some touristical POI, such as charging stations for ebikes, benches, picnic tables, public toilets and playgrounds. FOr this, a custom mapcomplete theme was created. (For a full explanation, see the last paragraph)

A part of the campaign involved a guided import. The agency had many datasets lying around (e.g. about benches or picnic tables) which they wanted to have imported in OSM. As doing a data import is hard and the data was sometimes outdated, we opted for a crowdsourced approach: for every possible feature, a map note was created containing a friendly explanation, information links, the tags to create and instructions to open MapComplete. When opened in mapcomplete, the user would be prompted to import the point or to mark it as not found or duplicate. All of these actions close the note with a small message on what the chosen action was.

Most map notes are closed by now, but the central question in this analysis today is: should remaining map notes be closed in batch, or do we leave them open for longer? Note that input of the local community will be gathered as well - this article will mostly serve as a point to start the discussion.

The datasets

Various datasets were provided to upload - which were converted into notes. In the table below, you’ll find a breakdown by topic, the date when they were uploaded, the number of notes created and how much of those notes were already closed and the top contributors for the category.

In this table, I’m not including if the feature has been added to OpenStreetMap, has been marked as not existing anymore or marked as being a duplicate.

Most of those notes have been opened by a dedicated account, except for two imports which accidentally did not use this account (noted in the table below).

Feature type (source) date Number of features Handled Handled percentage Contributor (closed notes)
Toilets (Toerisme Limburg) Notes created by Pieter Vander Vennet January 22 177 84 47% Eebie (26), Pieter Vander Vennet (11), phillipec (9)
Benches or maybe a picnic table (Toerisme Limburg) february 22 730 722 99% Eebie (378),A127 (90), joost schouppe (36), dentonny (34), jozin-belgium (32), mjans (23), Pieter Vander Vennet (13), kersentaart (13), proudtobeevi (11)
Playgrounds (Toerisme Vlaams-Brabant) march 22 51 46 90% Pieter Vander Vennet (23), Joost Schouppe (6)
Picnic tables (Toerisme Vlaams-Brabant) march 22 222 172 77% A127 (44), Eebie (15), Joost (15)
Ebike Charging station (Toerisme Vlaams-Brabant) march 22 20 13 65%  
Picnic Tables (Toerisme Oost-Vlaanderen) march ‘22 33 28 85% Joost Schouppe (10), A127 (5), L’Imaginaire (5)
Cycle rental (Toerisme West-Vlaanderen) april ‘22 5 5 100%  
Benches (Toerisme Antwerpen) april ‘22 54 44 81% pi11 (11), A127 (7)
Picnic table (Toerisme Antwerpen) april ‘22 91 73 80% Jakka (21), wegspotter (13), pi11 (12)
picnictable (Westtoer) Made by L’imaginaire april ‘22 340 296 87% L’imaginaire (250), Jakka (10)
Blue bike cycle rental (website scrape by Pieter Colpaert) may 22 60 44 73% Pieter Vander Vennet (12), Joost Schouppe (4)
Benches (Toerisme Vlaanderen Kempen&Maasland) may 22 94 88 94% Eebie (49), A127 (13), Joost Schouppe (8)
Benches (Open Data Oostende) june ‘22 1044 (!) 686 66% A127 (520), Joost Schouppe (41), L’imaginaire (36), Jakka (23)

Over time

How did the notes evolve over time? Are trends visible? The following graph shows the number of open notes for this campaign over time. The blueish line shows the total amount of Open Notes, which sharply jumps upwards when a new dataset was added.

Other lines represent the amount of notes closed by an individual contributor. As is visible, A127 and Eebie have done a tremendous amount of work, whereas around 20 other contributors have contributed a modest amount of points.

The amount of work represents a clear power curve, with most of the work done by a few contributors and many contributors with a few imports.

Another interesting graph is how much of the features got imported and how much got refused. As it turns out, 53% of all features got imported. Note that, if the point gets added by using mapcomplete, the note will be closed with the message ‘imported’. As many contributors used other editors, I’m checking for other keywords as well to mark them as ‘imported’, ‘duplicate’ or ‘not_found’. Of course, humans are messy and the keyword-based approach is incomplete and inexact. 13% of closed notes could not be matched automatically.

About 6% of the points were marked as ‘could not be found’. Some of the manually closed notes indicate that the area has changed and the feature (often a bench) is removed. As such, this is a good indicator or the staleness of the source dataset.

The fact that 6% of the features to import turned out not to exist anymore, this is a good argument for not blindly importing data into OSM! But this also poses that we should maintain the map and that features such as benches should be checked regularly. Ideally, the municipality administration would integrate updating OSM into their flow…

At last, even though no notes were created if a similar feature was already in OSM, about 10% of the notes was rejected as being a duplicate. This is partly because one dataset of benches turned out to also contain picnic tables - good for 168 ‘duplicate’ entries, yet duplicates are quite common in other datasets too.

Some other numbers

In total, 2921 notes have been created, of which 78% has been handled - that are 2301 that have been reviewed and imported (or closed with an indication that they cannot/should not be imported). Excluding the later benches of Oostende, only 7% (!) remains open.

That is a huge effort, of which I would like to thank all involved. Especially A127 who closed 684 notes and Eebie who handled 474 notes - your work is amazing!

The work of A127 is amazing by the sheer volume of the work, but I want to give Eebie an extra thanks as he took it upon himself to search for the toilets. These were notoriously hard to find and to survey, as het often tried to use the actual toilet in social facilities, group nursing homes or administrative centers. These were often closed or unaware that they were listed as having a public toilet; at other times, those toilets were marked >100m away from the actual location.

Differences between the datasets

The response on the datasets and the imports varied heavily by the type of the feature.

The benches and picnic tables are relatively straightforward. Visiting the place - physically or virtually with aerial imagery or Mapillary - suffices to decide if the feature still exists. As such, those tasks got handled relatively quickly. Only where no Mapillary and no aerial imagery are available, the map notes remain.

The other datasets proved to be harder. The playgrounds often needed some local knowledge - e.g. a playground might only be accessible to the members of the local youth organisation; or the administration eagerly labeled a patch of grass where kids could play some football as a proper playground. Some of them are hard to do remotely.

The hardest dataset to handle are the toilets, especially toilets in municipality buildings and social facilities. They cannot be seen on aerial imagery by definition, neither is Mapillary available. Furthermore, these facilities are often subject to opening hours and the rules about use by the public might change. In other words, a survey is necessary for pretty much every feature to import.

The reaction

We had a few negative reactions on creating this amount of notes. The arguments mostly boiled down to either:

  • new people will not bother to create a new note if the map is already littered with open notes
  • it breaks my workflow, I want to see and handle new notes.

For the second complaint, you can use this mapcomplete theme which shows notes and allows to filter them or create a new note.

I don’t agree with the first complaint as well, as the OpenStreetMap-website only shows a limited number of notes too. This can be easily seen by zooming out when the notes are open; you can see them disappear: https://imgur.com/a/PkwRe0h

Conclusions

In hindsight, the guided import was a success. By creating thousands of notes, the process was very discoverable and many people helped out, including two ‘hero importers’, but there are social limits to the amount of notes that can be created like this. For small datasets (<100 points in a single city), I would be tempted to create this kind of notes again. For bigger datasets (especially if >500 points), I’d probably opt to use MapRoulette to store the data and to mark it as ‘done’, as not to pollute the notes too much with such an import.

Especially small datasets which can be armchair-mapped have good levels of completion.

As there still is some activity on the notes, there is no reason to close them. On the other hand, there isn’t much activity anymore. As usual with this type of project, the last 10% is also the hardest to do and will probably take a long time before being handled. At the same time, there are only 79 notes remaining from the earliest import, so they don’t bother many people. (The benches-dataset of Oostende still has 334 open notes, about one third of the dataset.)

Please, discuss this on this forum thread. Remarks about the methodology or other musings are welcome here of course.

The project in a nutshell

This is what I sent to the Belgian Mailing list on the 5th of february. Note that the actual launch was later, around the 4th of March

Toerisme Vlaanderen will be launching an OpenStreetMap-based project on the 14th of February. This is a rather big project which I’d like to introduce to you with this email. The project consists of a few parts which might have some impact:

  • A MapComplete theme with focus on some touristical POI’s will be launched
  • A guided survey/data import will be started
  • Toerisme Vlaanderen will ask their partners to start mapping, so hopefully we’ll welcome a group of new mappers

This project will focus on the following POI’s:

  • Benches and picnic tables
  • (Public) toilets
  • Playgrounds,
  • electrical charging stations (with a focus on charging stations for electrical bicycles)
  • Bicycle pumps and repair stations
  • and observation towers

    *What is this project about?*
    

/Toerisme Vlaanderen/ is a Flemish state agency which promotes tourism in Flanders, together with the 5 provincial touristical offices and some other organisations. Historically, these 5 offices have each held their own small set of geodata for typical items such as benches, public toilets, picnic tables, playgrounds, … to put those items on their maps.

And of course, these offices have kept this data in their own spreadsheets, in their own formats (except for one - which has been using OSM for years now).

Toerisme Vlaanderen would like to unify all these databases into OpenStreetMap, increase the data quality of the items already there and improve the surveying flow.

This is where MapComplete comes in. MapComplete is a web-app where one can show information about POI, can answer questions about these POI and can add new points. Depending on the chosen map, some categories of POI are shown.

For this project, a theme showing (and asking information about) benches, picnictables, playgrounds, charging stations, … has been created and will be launched on 14 februari/. (If you look around a bit, you can already find a link to the theme, but another email will follow when the project is live.)/

  A slow data import: methodology

Of course, there is quite a bit of geodata laying around with the provincial offices, which ideally ends up in OpenStreetMap too.

For this, a slow data import has been started. Instead of dumping all the data into OSM, a map note is created for every item that should be checked.

This map note is structured in such a way that a contributor can use it, but MapComplete can also pick this up to show this to a contributor. This contributor can then quickly add/import the new feature if they found it, or they can mark the note as a duplicate or not existing anymore - closing the note in at the same time. These notes also link to both the wiki page and to the mapcomplete page where they can be easily added.

As with all things in life, this method has advantages and drawbacks:

  • The biggest advantage is exposure of the import to experienced contributors. (Case in point: I’ve already had valuable response on a few notes within 24hours of them going live)
  • The import is tracked in OSM itself, containing a lot of information and providing a flexible forum
  • It is quick and easy to setup
  • Others can make a similar note, which will be picked up by MapComplete too!

10 days ago, I wrote an essay about Bing Map Builder and how it could be used to fork the OSM community.

I made a prediction there:

Assume that Bing Map Builder becomes a really decent and good editor and that about a third of the edits happen through Map Builder. Microsoft could then -at some time in the future- decide to let updates from Map Builder flow to Bing Maps first, and only let them flow towards OpenStreetMap at a later time, “to review them for quality”.

It seems that this prediction has become true already (1). In the discussion under my previous entry, people noticed that “no bing accounts appeared anymore” in the new to OSM-listing. Time to re-investigate!

So, what is the behaviour now? I drew a new building, clicked saved and… the building disappeared from my screen. When opening the network console, this network call proved my suspicions. The created data is now sent towards https://bing.com/mapbuilder/changeset/submit and contains the changeset data (and bit of extra information)

Changes made with Bing map builder are now sent to Bing, not to OpenStreetMap anymore.

(Edit 2023-03-01: Allison P pointed out in the comments that the building she drew now appeared into OSM, specifically in this changeset)

I’m curious what caused this and how this will play out further.

What caused this?

One hypothesis is that my essay caused them to take action and to switch over to their private backend to mitigate my critiques.

On the other hand, this might have been planned all along and have been rolled out just as planned.

Or maybe the truth is in the middle. It could have been planned, but rolled out sooner because of my essay.

In either case, I wasn’t contacted by their team - so I don’t know. Maybe I’ll hear how this went one day; maybe not.

Does this solve the issues I complained about?

The upside is that the community does not have to deal with pseudonymous accounts anymore. And the product “Bing Map Builder” now at least does what it says on the box.

The drawback is that this fractures the community further, just like I feared.

And the attribution to OpenStreetMap is still missing. As long as OSM-data is shown in Bing Map Builder, a clear attribution should be shown somewhere.

But the email delivery is fixed though. The first email I received arrived about half an hour after I published my article.

How will this play out further?

I’m wondering how this will evolve further. Will Bing upload the data that their contributors to OpenStreetMap - after a review and conflation? Or are we seeing a de-facto fork between the two datasets? Time will tell.

Some personal experiences about publishing the essay

My essay did make quite a splash and was shared widely. It got me on the frontpage of HackerNews (achievement unlocked) and my Mastodon account gained about 100 followers. The essay got translated and republished in French.

But most importantly, I got many positive reactions, saying that they “enjoyed the read”, found it “an interesting perspective”. Some wanted to take action by “reporting the Bing Map Builder account” (which has been blocked for quite a while), others found I was a bit “alarmist”. However, nearly everyone who commented agreed with my talking points. And I got no (very) negative reactions (e.g. no rude reactions or ad-hominems) - all messages were polite.

I didn’t receive official communication from neither the OSMF-board nor the Bing Map Builder team (yet?) - but a few OSMF-board members reacted via various channels. I trust that they are following up on this - I’m quite sure that the OSMF communicates with the Map Builder Team regularly to follow up the project and that this communication was ongoing before my essay. (For clarity, I had no contact with OSMF-members about the content of both essays as well).

Footnotes

(1) To be precise: the last part or this prediction has become true.

OpenStreetMap is in trouble

Posted by Pieter Vander Vennet on 6 February 2023 in English. Last updated on 20 October 2023.

Edit: A small update happened after publishing this article - you can read it here: https://www.openstreetmap.org/user/Pieter%20Vander%20Vennet/diary/400992

OpenStreetMap is in trouble

It is a long-standing tradition that every now and then, a member of the OpenStreetMap-community posts that OSM is in trouble. Often times, these essays complain about some trivial things which are, in the end, not that important. For example, they complain that we didn’t implement Bézier curves yet (we don’t need them), or that the data model is stale (it isn’t, new tagging appears every day), that the main OSM.org website doesn’t have some feature and isn’t on par with Google Maps (that’s by intent) or that AI will make the entire manual mapping space obsolete, in “just another ten years time”.

However, most of these things miss the crucial point of what OSM is: a community; a group of people that are working together on mapping the world in an Open Data way and building related tools with Open Source. Our strength is the unison in this goal, even though everyone pursues this differently, through different technological means and for different motivations. Motivations range from the most mundane reasons up till political activism. And that’s fine. All this activity and diversity strengthens us as a global community.

However, recently, a new participant has entered the ecosystem with parasitic intents. It tries to capture away precisely what makes OSM strong: the contributors.

The means to this end is called ‘Bing Map Builder’.

A bit of history

As you all are aware, OpenStreetMap-data is republished under the Open Database Licence. This means that everyone can use OSM-data for all purposes (including commercial purposes), but they have to honour two obligations:

  • You have to attribute OSM
  • Changes to the data have to be shared again under the same licence

I’m glossing over some of the technical details, but the spirit is that the licence is contagious. I argue that this is one of the big reasons that OSM has become the successful project in this space, and I argue that two other important projects are successful thanks to a similar contagious licence: Wikipedia uses Creative Commons with Attribution and Share-alike, the Linux Kernel uses the GNU General Public Licence which is share alike similarly.

All this serves a specific purpose: it binds people using the data or tools to contribute back their little bit of help. On an individual scale, this contribution is negligible; yet all those little drops become a huge river together. These licences force us to unite under a common banner, to put the community above the individual.

I’m not blind to some of the drawbacks that this approach has. I’ve been in many meetings with interested parties where this licence proved to be puzzling to the uninitiated or where the share-alike licence was outright incompatible with their goals; even though these goals were to publish open data from the government! The share-alike-license is often incompatible with the obligated public-domain-license they sometimes have to use.

Yet, I’m still convinced that using a CC0-license for OSM would kill the project instantly. Consider, hypothetically, that every contributor who ever made a change would agree to convert all their previous contributions to CC0, what would happen (1)?

A data run would happen. Every actor out there, big and small, would grab a copy of OSM and go their own way with it. I’ve seen countless similar projects, but all focused on a small scope, all in varying states, such as Open Benches, Camper sites, Slipways, charging stations AEDs, and then I’m not mentioning the thousands of maps on municipality websites.

All of them would grab a partial copy of OSM with only the data that they are interested in and try to wall them off and build their own community around it, effectively splintering OSM-contributors into thousands of niche communities.

The result would be that more work should be done in total, for way worse data quality as the data would not be integrated any more. Tooling and software would become just as fragmented, killing the quality of the software too.

So, even though the direct move of using CC0 for the OSM-data can be considered “more open”, it would in the end be a big net negative for Open Data.

The biggest victims of this move would be the OSM-based general purpose consumer applications (OsmAnd, Maps.me, Organic Maps, …). They don’t have a specialized community around them, and they would see their road network and POI data going stale, slowly forcing them out of the market.

The biggest winners would be the closed source mapping applications run by big corpos, such as Google Maps, Apple Maps and Bing Maps. They have access to resources we can only dream of. Where we had to do with a volunteer sysadmin for over a decade (2), setting up a map database is not that difficult for such a big organization who can hire a team right away. They can buy basic datasets from specialized companies without much ado, and they have access to people via their respective devices. These devices can help to build the map data via prompting the user (“what is the phone number of the restaurant you just visited?”) or by spying on them (e.g. movement data to build road networks, max speeds, and one ways) - which gives them a big advantage.

A confession

A bit of a sidenote: it is not necessarily bad to have a subcommunity within OSM which is focused on a specific subject (e.g. trees, or artworks, or cycling related stuff) - as long as they all play along nicely with the wider ecosystem, that is a huge bonus. Even more so (and this is a disclaimer of my huge bias): as developer of MapComplete, I’m building precisely those tools to give such a subcommunity the means to focus on just a single topic. However, under the surface, a lot of the code tries to make sure that those tools integrate well with the data by trying to prevent duplicates or make links to other themes. And the interface also guides new contributors to discover the other topics and other editors so that they can grow into a better contributor and community member as well.

Another confession: at my job at Anyways, we have done precisely this fragmenting of data, by creating a small map where only a few people can edit. But before you grab the pitchforks: those sandboxed (and private) copies of OSM serve to study possible transportation changes in the city. The researcher makes hypothetical changes (such as closing a street for through traffic or building a new cyclist bridge) and can then see how cars/cyclists/pedestrians would behave in the new scenario. As such, this data is not suited for OSM in the first place. Even better, as we reuse the iD-editor, those researchers gain OSM-experience and often update OSM themselves once the mobility changes go into effect.

The current situation

To get back to reality, we see that Google Maps is a runaway consumer success, partly due to the gigantic access that google has to people’s daily lives and due to the deep pockets that google has to fund this all. Furthermore, thanks to the coupled sales of the Play Store, Google Maps is pre-installed on every sold Android phone which drives hoards of people to use Google Maps.

Several other big companies - most notably Apple, but also Microsoft - have tried to emulate this, without much success. The non-google platform companies have a weak spot there. On one hand, they really need a map to offer a feature-complete device (and to get access to that sweet geo-data of consumers). They can’t rely on Google Maps by Alphabet Inc. - it’s their biggest competitor. If they did a price hike or cut access to their services, they would be severely crippled. At the same time, keeping a world-wide up-to-date map costs billions of dollars annually - a hefty sum to pay alone. And yet, they are all so envious of Google Maps. They all want to have it, but only one can be the top contender.

And then there is this weird (3) project that gather extremely qualitatively data on a big scale and gives it away, for free. Why would anyone do so much work without asking money for it?

But even better would be that all those people would do this work for free, but in a way that contributes to the bottom line of the corporation. In the end, many people contribute to Google Maps, Waze and quite some other projects where they sign away their rights to that company. And let us face it: Microsoft would love to get a contributor base for bing maps - and they are working towards this goal, piggybacking on the efforts of OpenStreetMap. Bing Map Builder is a first attempt to do this.

What does Map-Builder do (on the surface)?

The map-builder is discoverable through bing maps if you are browsing around in Australia. If you are browsing around there, an extra button appears in the bottom right with “edit maps” just above the ‘Feedback’-button. Additionally, a big banner is shown on top which invites people to make map edits.

Once opened, you are greeted by a modified iD editor which has a nice tutorial which invites drawing a building, a road or a water body (4).

This sometimes goes wrong horribly. In this case, the map builder didn’t realize that the water body was already there. I started improving the geometry of it, after which I was completely unable to actual save and upload this… The ‘save’ button correctly realized that there was one change, but I couldn’t click it. Unselecting the feature gave a warning with the choice to ‘continue tracing’ or to ‘exit’. Continue tracing did take me back to the broken state and clicking ‘exit’ resulted in my changes being erased.

The last important feature shows messages and changeset comments in the editor itself. It is a nice feature which I would like to see upstreamed in the editor. The reason that this button has been added, however, is due to dark reasons.

What does MapBuilder actually do?

It edits bing maps! Right?

Have a good look at the previous screenshots. Something that is completely missing, unless you go and search for it very deeply. Do you know what is missing?

…. Take your time, scrutinize the screenshots ….

In case you haven’t realized yet: there is no mention of OpenStreetMap in sight.

It is mentioned somewhat if you open up the tutorial page and there is a ‘view on OpenStreetMap button’ if you open up the sidebar, hidden away in the top left.

In the same vein, there is no clear step where contributors have to agree to the contributor terms. This is the view you get when signing in:

For those which don’t have an excellent eyesight - this is the text at the bottom (with links expanded):

Your edits will appear in Bing Maps and OpenStreetMap (*). Edits and comments will appear publicly, but your profile information will not be shared. By proceeding, you agree to OpenStreetMap Community Guidelines (https://wiki.osmfoundation.org/wiki/Licence/Community_Guidelines) and Terms of Use (https://wiki.osmfoundation.org/wiki/Terms_of_Use).

((*): Clicking ‘OpenStreetMap’ will open up the help panel, of which the screenshot is just above this screenshot)

Only if the potential contributor spots this legal text, clicks on the ‘Terms of Use’ and continues towards the ‘contributors agreement’, they can learn under what license their contributions will be published. (11) It took me two accounts and actively seeking this out before I noticed them.

In other words, 99% of the users will never see this. 99.99% will never realize that they are editing OSM. This is not only in breach of the letter of the ODbL (and thus illegal), this flies into the face of everything what the OSM-community stands for. It strongly suggests that Microsoft actually wants: people contributing to Bing, not to OpenStreetMap.

It protects privacy! Microsoft really cares about that, right?

This reach extends in the way that accounts are structured. Signing up via map builder creates an account on OpenStreetMap which is a bunch of ten random characters such as the account created for my testing. Officially, this is “to protect the privacy of the mapper”. There is some merit in this. As a big company, you can’t publish personal data of your users (12). But you can ask for their consent and then publish it. You can give them a choice. I tried to take control of the account listed above by logging in to OSM with the Microsoft Account used for the account above, after which OSM told me that “No OpenStreetMap account is associated with this email address”.

There are many issues with this approach:

  • This is confusing, as this means that I’ll have two OpenStreetMap-accounts associated with the same email-address.
  • It takes away all personality of the mapper. The faces and people are hidden away behind a random string, without even the possibility to state who you are…
  • … and precisely having an identity is central to having a well-functioning community. Trust can only grow by having multiple interactions with people and us remembering these interactions. Human brains are terrible in remembering random strings - I can’t even remember my own MapBuilder-username on OSM. Remembering hundreds of such strings will be completely impossible.
  • Furthermore, by having accounts managed by Microsoft, this takes away the control from the OpenStreetMap-admins. It’ll be nearly impossible for the DWG to moderate this type of accounts once that malevolent actors discover mapbuilder, e.g. as a tool to produce spam on OSM.

But the community will be able to contact the mappers in case of trouble, right?

Another feature of map builder is the ‘notifications’ area in the top right corner. It shows changeset messages and personal messages in a ‘chat’-like interview. I have to admit, this is one of the few features which I really like from a usability point of view:

But something that is missing in this view is the possibility to send messages to other community members that haven’t contacted you before. This takes away another fundamental ability of the community, namely the ability to connect and the ability to organize. At last, providing feedback and organizing is important to build a community (and a good map).

Furthermore, there is no possibility to inspect the history of an object. At last, there is no way to discover other means of communication that local communities use - in contrast to iD which nicely shows the community index when uploading is done.

This takes away the possibility to be a human being to each other, especially when humans are not even allowed to have a name.

And to make things worse, this is actually a setting:

EDIT: about half an hour after I published this essay, emails started arriving when a message is sent. The next paragraph is thus not relevant anymore: One can turn off email notifications. I turned them on, but I’m really wondering why the inbox of the Outlook account looks like this (9) - despite having sent multiple OSM-messages to this account:

We had situations in the Belgian community in the past where a very active mapper made a very annoying systematic error. Changeset comments and messages did not help at all. However, some people from the local community knew him in person - giving us a side channel to send him a friendly worded but firm email of all things that had gone wrong. Within hours, he showed up within our chat channels apologizing. It turned out that his OSM account was tied to an old email address that he didn’t check any more, thus all the effort we put into changeset comments and sending messages went to waste (8).

And I’m not done yet.

Again, at first sight, it is great to send messages to other users. I would love to have something like this in MapComplete. However, I can’t reasonably do this. There is no API to read the user messages, let alone to send a reply - due to concerns for spam messages. Of course, I could set up a way to technically achieve this: when someone logs in, I can send their access token over to my server and then scrape the messages from the website. If I were to do this, however, I would face backlash in several ways:

  1. Sending over access tokens is a no-go in terms of security.
  2. I’ll end up with one server which has many login-credentials, a huge job to keep safe and a ticking time bomb which I do not want to manage (I’m just one person and would be personally responsible).
  3. And the OSM admins would quickly block that IP address as it might be mistaken for a spam channel and because it is a security risk.

Due to the size of Microsoft backing this project, they could build a feature that all other editing programs cannot build. This is another attempt to shift power away from individual contributors and other project towards Microsoft, which effectively becomes a man in the middle between the community and an individual mapper.

Every account is created equal, right?

As it is not possible to login into OSM with these accounts, this also means that a contributor cannot choose which editor they use. An account created with Bing Map builder cannot be used to edit with iD, JOSM, Vespucci, StreetComplete or MapComplete. This too is unprecedented - as this means that we now have two classes of OpenStreetMap-accounts (5), of which one category is vastly more restricted. This class of slave accounts cannot choose which editor they use, they cannot build an identity within the wider OSM-community. They cannot even effectively communicate with the wider community, nor can the community effectively give them feedback.

In other words, contributing via Bing Maps denies the contributor their basic humanity, it isolates them from the community and limits them severely in what they can map. And it is detrimental to the long-term health of the OpenStreetMap community.

Embrace, Extend, Extinguish?

We should not forget history. Even though Microsoft did contribute a few things to open source and donated e.g. Bing Imagery (10), it is still a company which primarily wants to make money. And even if these actions are well-intended, we should judge them by the effects that they have or could have.

Let us speculate about a possible future.

Assume that Bing Map Builder becomes a really decent and good editor (7), and that about a third of the edits happen through Map Builder. Microsoft could then -at some time in the future- decide to let updates from Map Builder flow to Bing Maps first, and only let them flow towards OpenStreetMap at a later time, “to review them for quality”. After a while, these updates might then stop working too. Due to the ODbL, OSM would still be allowed to copy changes from Bing into OSM, but this will become harder as the Bing-fork drifts away - especially if they import e.g. the Buildings Footprint Dataset or other datasets. The difficulties will not only be technical (e.g. how to conflate different data versions of the same building) but also social, as a part of the community will refuse this sync for ideological or other reasons. (6). To complete the cycle, Microsoft might, at an even later point decide that all new contributions to Bing Maps use another licence or ask the contributors to also sign away the right of previous contributions they have made. Then, newly added features would have a different license and would be legally impossible to import back into OSM.

This will effectively result in a fork of the OSM-project - as both the data and two parts of the communities will start to drift away from each other.

What now?

We should not accept this behaviour by Microsoft. I’m calling to the OSMF to take up their responsibility and to get in touch with Microsoft to rectify these issues. I’m sure this is a big misunderstanding on my part and I’m overly pessimistic and that the team from Map Builder will quickly fix the following:

First: the attribution must be fixed. A clear text should explain contributors that they are joining the OpenStreetMap-community, not the Bing-maps-community and that indicates their data will be republished under the ODbL. A clear attribution should be visible throughout using Map Builder. This is non-negotiable as is explained in our copyright notice.

Second, the email delivery should be fixed. I did not receive any emails during my tests. Furthermore, the ability not to receive emails from OSM-members has to disappear as well. This too is non-negotiable. If the community cannot correct mistakes by beginners, then the map quality will plummet.

Third, people should be able to log in with their Microsoft-created account into OSM to take control of it, e.g. to use a human username and to introduce themself through their profile. Reversely, contributors with a normal OSM-account should be able to log in into Map Builder.

I’m proposing that the OSMF enacts an ‘editor neutrality’ in some way, that every OSM account should be able to use every other editor that is out there and that no third party ever can take control over user accounts in such a suffocating way. Even stronger, I’m proposing that every program that makes changes to OpenStreetMap via the OSM-api should be Open Source - to protect the fundamental freedoms of the OpenStreetMap community.

If no improvements at all happen, we should consider cutting off mapbuilder from the API, and we should consider starting a lawsuit over the copyright infringement. This is not a small hobby website which happened to have forgotten the attribution. This is a cancer that is starting to grow. This is designed to kill the community.

Footnotes:

  1. The process of trying to convince everyone to switch to CC0 would in itself be highly contentious and split the community; probably causing a fork and effectively killing OSM as well. The community has experienced such licence shifts - even though I wasn’t around back then.
  2. Thank you for doing this humble work years on end, Grant Slater and Tom ! You are a heroes! And not to discount your work, but having a volunteer running such an operation is - from the perspective of the organization - a weakness. A volunteer might quit, have other stuff becoming more important in their life, might burn out, …
  3. Weird from a capitalist view of the world
  4. As the tutorial invites to draw a building/waterbody where some dataset thinks there is one, doesn’t that amount to a data import? It might be executed in a distributed way by newbies, but IMHO this is still a data import. I’m also wondering what the source data set is and how it is licensed?
  5. To be precise, there are a few privileged accounts which can do user blocks and retract data. They are, however, managed by people trusted in the community to do the thankless work of cleaning up all the shitty cases, which can be contentious. A big thanks to all the members of the DWG for this!
  6. Whether I would be opposed or in favour of such a sync is totally irrelevant. We already know that a part of the community will be in favour and a part will be against - and this contentiousness situation is what we must avoid.
  7. Probably not going to happen though - but we should still oppose this.
  8. Afterwards, this person stayed an active contributor and often helped out
  9. This is a lie - I did delete four messages from this outlook-account: one welcome message and three “advertisements” which looked suspiciously like predatory visa card scams. They turned up right off the bat, even though it was a fresh account only used for OSM. Well-intentioned messages don’t arrive; but advertisements preying on the weakest in society do.
  10. And Bing imagery is a poisoned gift as well. The technical requirements to implement it involve getting a token for every user using the map, making it trivial to track what areas an individual looked at - even in other map applications.
  11. I strongly question if OSM could make the case before courts that a contributor was well-informed about the license that they agreed to (even though I’m sure that such a scenario will not happen)
  12. But they don’t mind collecting personal data and use that for advertising7

Is all corporate involvement evil?

At last, as closing note: not all involvement from big companies into OSM is necessarily evil. Often, our goals will just happen to align (even though the means will cause some friction).

Take, for example, Amazon. They added tons of driveways to OSM to be able to make deliveries. There is nothing evil about that, and I even encourage this. However, their means - paid labour - sometimes causes friction with volunteer mappers. If they onboard 100 paid mappers, some of them might not have paid attention during formation or make a systematic error before it is corrected by a volunteer mapper, causing frustrations. The goals might thus align, but as execution of it is not perfect, this might cause conflicts and some harm. As one of the actors is a big corporation, there is a huge asymmetry in the power dynamic - and they should be held accountable if something goes wrong - but I’d still consider this as an absolute win for OSM.

Text licensed under CC-BY-SA Feel free to translate this, but link back to this page and mention my Mastodon: https://en.osm.town/@pietervdvn

An overview of reviews made with MapComplete

Posted by Pieter Vander Vennet on 29 January 2023 in English. Last updated on 31 December 2023.

MapComplete has - for some thematic maps - the ability to leave a review on an entity with Mangrove.Reviews. Up till now, I had no idea how much this feature was used. However, due to technical reasons I had another look to the reviews module and discovered the ‘download all’-option on mangrove.reviews

Mangrove Clients

The analysis was made with data from 20 january 2023, downloaded around 17:00 UTC time.

This data contained 660 reviews. As the website making the review is recorded, we can make a breakdown of the top websites:

  • https://mangrove.reviews is unsuprisingly the most popular website to make reviews on, with 318 reviews made
  • MapComplete is the second (and the biggest ‘external’ website), with 192 reviews (of which 13 are made with the development version)
  • toggenburg.swiss is third, with 35 reviews

A variety of smaller websites follows, each with a few reviews made. At first glance, most of them seem to be swiss or german. Furthermore, there are 5 reviews made by localhost:1234 and 7 by localhost:5000. The former is probably me, testing the creation of reviews while developing.

The full table is listed below.

Client website Number of reviews
https://mangrove.reviews 318
https://mapcomplete.osm.be 179
https://toggenburg.swiss 35
https://thurgau-bodensee.ch 29
https://pietervdvn.github.io 13
https://open-reviews.orderingstack.com 13
https://heidiland.com 12
https://open-reviews.net 12
https://winterthur.com 9
https://ostschweiz.ch 7
https://localhost:5000 7
https://actuallyfoods.com 5
https://staging.mangrove.reviews 5
https://localhost:1234 5
https://examples.twblue.cms.tso.ch 3
https://ivydog.art (has NSFW content) 2
https://toggenburg-aktiv.ch 1
https://modernmount.blogspot.com 1

Themes

It seems that pet owners gave reviews the most. The ‘pets’-theme - first known as the ‘dog’ theme - yielded 35 reviews in total (if summed). The ‘Pin je Punt’-campagin by Visit Flanders (which included pubs and cafés) is a close second with 29 reviews, followed by the shops theme.

While it might be surprising that the pets-theme has such a high number of reviews, the reason for this is straight-forward: the review-module is placed on top of the infobox for vets, whereas most other layers do have the images module there.

Theme Review count
https://mapcomplete.osm.be/pets (incl 14 reviews made under the previous name ‘dog’) 35
https://mapcomplete.osm.be/toerisme_vlaanderen 29
https://mapcomplete.osm.be/shops 28
https://mapcomplete.osm.be/food 18
https://mapcomplete.osm.be/playgrounds 16
https://mapcomplete.osm.be/cafes_and_pubs 14
https://mapcomplete.osm.be/personal 14
https://mapcomplete.osm.be/campersite 9
https://mapcomplete.osm.be/sport_pitches 9
https://mapcomplete.osm.be/fritures 8
https://mapcomplete.osm.be/climbing 7
https://mapcomplete.osm.be/onwheels 3
https://mapcomplete.osm.be/hackerspaces 1
lgbtmap (theme on a fork) 1

Language

At last, due to the full client URL being saved, we also know what language the user interface was in when the review was made:

language count
en 77
nl 37
de 27
ru 13
fr 9
da 6
es 2
unkown 21

Locations

And as map makers, a quick geolocation-analysis can’t miss as well.

Unsurprisingly, most of the reviews are in Europe:

All reviews

When zooming in, a clear cluster in Switzerland and Flanders is visible:

All reviews in europe

When keeping only the Reviews made with MapComplete, the cluster in Flanders remains:

All reviews in europe made with MapComplete

(Copyright: pins are Mangrove Reviews, background map is OpenStreetMap of course)

Conclusions

So, no big conclusions and surprises in this little analysis. The reviews are being used - but not very much, especially in comparison with the image-upload feature. Placing the reviews higher up might help or including in into the questions flow. The core question is how important reviews are within the vision of MapComplete.

As usual, data is available online here, script is here

What licenses are used?

Now that MapComplete is two-and-a-half year old, it’s a good time to see what license people are using to upload their images.

Why do I care?

The first reason to do this research is curiosity. How much pictures are uploaded with what license?

The second reason is a very practical and UX-driven: if a significant portion of contributors doesn’t bother to change the license, then the license picker can be moved from the ‘infobox’ into the ‘user settings’, freeing up valuable space there. User tests have pointed out that this is valuable.

Methodology

MapComplete uploads images to imgur.com and then links to this image using image=https://i.imgur.com/aBcDeF123.jpg. Some metadata (most notably the author and chosen license) is added as ‘description’ to the image on Imgur. If multiple images are added, then keys image:0, image:1, image:2… is used.

At last, themes can also add images under a specific key. For now, only the etymology-map does this with image:streetsign.

Overpass was used to download all features with a tag matching one of the described keys and matching an imgur-url.

Then, the description of all those images is downloaded and parsed, yielding the needed metadata.

Even though some people did add images to imgur to link them to OpenStreetMap before, we assume that (nearly) no images will also have the license information encoded as MapComplete does. Furthermore, this does not keep images of now-deleted features into account, nor does it take images into account that have been deleted in the mean time. I don’t think it’ll make a big difference though.

The resulting datasets are here. The script to download this all is in the MapComplete repository. Keep in mind that using this script will exhaust the daily IMGUR rate limit; so please use a different access token or spread the download over two days as was done for this research.

Results

In total, 12516 images with a parsable license were found - this is a huge amount of pictures, which I did not expect! This was done by 439 contributors in total

Unsurpisingly, the vast majority was uploaded with the default license, being CC0/public domain. This is about 10635 total pictures (or 84.9% of all pictures), taken by precisely 400 different contributors - 91.1% of contributors.

The second most popular license is the creative commons with attribution and sharealike license (CC-BY-SA), with 1707 images in total, or about 13.6% of all images. However, only 32 authors choose this license, or 7.2% of the photographers. Striking is that those are way more active, with an average of 53 images/person!

At last, the creative commons with attribution (CC-BY) is not popular at all. Only 117 pictures in total - 0.9% of all pictures - used this license. Only 10 authors picked this option, which also indicates that they are below-average in number of pictures taken with 11 images/contributor.

When the authors which used CC-BY and CC-BY-SA are summed, only 42 are found. This indicates that there is a big overlap between contributors that used the CC-BY license. Personally, I contributed under CC0 first, then a bit under CC-BY to switch to CC-BY-SA for the most part of my pictures. Other contributors probably did a similar trajectory.

Oh, and due to a bug, the license of some images got saved as "undefined" instead of the actual license. This bug only impacted 57 pictures (0.4% of all) taken by 20 authors. As we don’t know the license they took, we should stick to the most restrictive of the available licenses to reuse those images.

Averages and medians

On average, a contributor with at least one image, makes about half, namely 28.5 pictures/person! However, this is a typical power curve, with a few powerhouses that add tons of images. The median contributor with at least one image contributes two images.

Conclusion

First of all, I’m absolutely flabbergasted by the total amount of pictures taken! I knew it had to be in the thousands, but never realised it would be over 10k!

As only 42 contributors ever contributed under a different license, I feel comfortable to move the license-picker away into the user settings panel. Freeing this place will improve the experience of thousands of people at the cost of a few clicks that only a handful of people have to make - even though that this handful of people are the most active contributors.

I’m also very positively surprised by the high number of average pictures per person - even though the median is a bit more modest.

And the fact that someone has uploaded twice as much pictures then I did is really cool to. It’s also the only contributor (so far) to go over 1000 pictures and is even getting close to breaking the 2000-boundary… Congratulations, Awo!

The second place is for me (Pieter Vander Vennet), with 859 pictures added. (Damn, this much already?)

The third place is for vjyblauw, another power mapper in Belgium with 746 pictures. Congratulations as well!

At last: I’ve attachted the top 50 of contributors below.

But before showing it to you, I’d like to tell you all one more thing:

Thank you for contributing!

This wouldn’t be possible without all of you

Position Username Total number of pictures
1 Awo 1953
2 Pieter Vander Vennet 859
3 vjyblauw 746
4 JLZIMMERMANN 645
5 Thierry1030 622
6 L’imaginaire 589
7 Jose Luis Infante 575
8 Toni Serra 446
9 APneunzehn74 439
10 joost schouppe 310
11 Maarten O 301
12 5R-MFT 254
13 Wolfram Hoppe 250
14 Koen Rijnsent 234
15 WimBau 229
16 dentonny 212
17 Stijn Matthys 137
18 Polardfront 126
19 TauvicR 119
20 Locatus_Jori 109
21 Locatus_Raf 100
22 Robin van der Linde 98
23 wjtje 88
24 Marival 75
25 Pieter Nuytinck 71
26 Vincent Bombaerts 68
27 Rober castro 65
28 349499 58
29 Frans_Napaters 57
30 Thibaultmol 57
31 philippec 56
32 StefDeGreef 52
33 borgofumo 52
34 ClarissaWAM 48
35 jospyck 48
36 escobrice 44
37 KaiPankrath 43
38 Ninopiña10 43
39 Niels Elgaard Larsen 42
40 RodrigoKiger 41
41 MAGONA 39
42 sjokomoeske 37
43 ccasado 36
44 Piotr Barczak 34
45 lololailo 34
46 Manuel C Arco Martos 33
47 reginaldc 33
48 Hilde OSM 32
49 paunofu 32
50 Gruppe 24(2) 30

Hi all,

Hacktoberfest is a yearly event where contributors get a T-shirt from hacktoberfest.com if they improve an Open Source Project.

MapComplete is open for such improvements. Head over to the repository. An ideal to get started is by creating a map layer about something that is interesting to you.

If you have questions on getting started, feel free to ask in our chat channel on matrix/element (telegram bridge )

Detect tree species automatically with PlantNet

Posted by Pieter Vander Vennet on 14 September 2022 in English. Last updated on 15 September 2022.

PlantNet.org offers an API which tries to determine the plant species based on some pictures.

I integrated this API into MapComplete, together with some queries to Wikidata. As a result, linking a tree to the correct species identifier is easier then ever and can be done with a few clicks:

Hey all,

In the past month, MapComplete was part of Open Summer of Code where 4 students and myself did make a lot of improvements and a new theme.

With this diary entry, I’d like to give you some insights in what we’ve done the past 4 weeks.

What is Open Summer of Code

Open Summer of Code (or OSOC) is a programme organized by Open Knowledge Belgium, which is a small belgian NGO that promotes Open Source and Open Data.

For OSOC, they search clients (organizations or governmental institutions) which have an interesting problem that they want solved and with budget to pay a team of about 4 students.

These 4 students will be guided by a coach (such as me) to make sure something useful comes out of it.

The actual problems are varied. We’ve had a planning tool for building new homes, a calendar application based on SOLID, a tool to discover research papers, …
The bottom line is that data must be open and that all produced software and tools will be open sourced. If possible, the programs should reuse existing tools and datasets, such as OpenStreetMap or wikidata.

OnWheels: the wheelchair accessibility map

One of the projects this year was paid for by BOSA (a belgian gov organization) requested by OnWheels - a belgian application which helps wheelchair users to navigate the world. They have a database of shops, restaurants and other amenties together with some info about them, such as name, contact details and opening hours, but also information about the width of the door, the height of the kurb at the entrance, …

During the past years, the idea of opening this data has grown within OnWheels, for various reasons. By opening the data, more people can reuse it. Furthermore, by switching to OSM, the cost of maintaining this data is shared amongst more people.

However, making the switch is not easy. With the OSOC-project, we wanted to create a first version of how an OnWheels 2.0 might work.

Whom is this app for?

During our scoping sessions, we identified three groups of people who will be interacting with this app:

  1. People with reduced mobility
  2. Data contributors
  3. Users which might export or analyze the data

People with reduced mobility

The first group of users is the obvious group - they are the main group of users and should be able to get the information they need quickly and easily. They include wheelchair users, but many more groups of people are served by having good data; including people without a disability. In a way, people with e.g. a stroller can benefit from this information as well.

Data contributors

The second group of people are the data contributors. They include the casual mapper, the group of people who do a mapathonn (e.g. as a teambuilding) to a municipality who has a dataset lying around that they want to import.

Data reusers

The last group are the data reusers, whom want to reuse and analyze the data. A typical example here is a government or municipality who wants to create a report about the accessibility in their city.

What is good wheelchair accessibility data?

OSM has a long tradition of mapping wheelchair-accessibility with wheelchair=*. As it turns out, this is rather limited. Some wheelchairs are wider then others, some wheelchairs are pretty long (e.g. wheelchairs with a third wheel in front; electric scooters often used by old people, …).

Some wheelchair users can cross big kerbs (e.g. by getting up, stepping over, lifting their wheelchair over the kerb and sitting down again), whereas other people with reduced mobility might not be able to cross a kerb of even 2 cm.

In other words, we want more detailed information!

This also raises an important data question. Should we add this information on the POI of the shop, on the enclosing building or on the door? We decided to add the information on the individual door objects. For example, the shop might move, but the door will stay for the next shop. Furthermore, a building (and thus a shop) might have multiple entrances with different properties… By keeping this information on the doors, the data model stays pretty clean.

Mapathon/immersive session

We also tested this in real life, by going out with the team in wheelchairs for a stroll through the city. Quite an adventure, especially taking the subway… You can find a movie about it on the project site.

The features

With these three groups of users in mind, we set out to create an application which served all of them.

Of course, MapComplete already goes a long way serving these groups. The main focus lay thus on creating a mapcomplete-theme for wheelchair users, with a few extra needed features on the sides.

Layer: Entrances

The first important layer where we all deal with are entrances. For wheelchair users, they are very important: if the entrance is too small or the kurb to high, it becomes an insurmountable obstacle.

Indoor navigation and level selector

The first major, new feature is the level selector. When there are features in view over multiple levels, an elevator will appear on the right where users can select which level they want to view.

In tandem, a layer showing indoor mapping features (such as rooms and corridors) was developed.

When mixing the indoors layer, entrances, walls_and_buildings and pedestrian_paths-layers, we get a basic indoor viewer. A good testing ground is the building we were at during the project.

_NB: this indoor viewer replaces the previous theme ‘entrances’

A dash of magic

Having an indoor viewer is cool, but doing some automatic analysis of the data makes all the difference.

The walls_and_buildings received an update. Every building-feature will calculate which entrance-objects are located within or on the edge of the building. These can be neatly shown on the building, giving an neat overview:

Layers: Toilets, reception desks, elevators and some more layers

Other important features for wheelchair users are toilets, reception desks and elevators. The toilet layer received some extra questions to gather data relevant for people with reduced mobility; and a layer with reception desks and elevators was added.

The elevator layer asks for information about the physical size of the elevator and offers the possibility to mark an elevator as broken or closed which, sadly, is often the case as we noticed during our wheelchair trip .

The kurbs-theme got some improvements too.

At last, a hotel layer was added as well to have some feature parity.

Features for data reusers

Last year, a feature to download the current view (as geojson or CSV) was already added (even though this feature is disabled by default - it can be enabled by adding an URL-parameter or by enabling it in the theme config).

For the professional data reusers, a ‘dashboard’ was added as well, where some stats can be seen in the blink of an eye.

Importing data

To import data, a layer to read maproullette tasks has been created which works similar to the import notes. More information on that will come soon.

Mixing it together

All these layers and features are mixed together in the OnWheels theme

More magic

At last, the onwheels-theme also ‘steals’ the doors-overview from the enclosing building-layer. Clicking a feature within a building with doors, will reveal the information about those doors.

More info

More info on the project page: https://osoc22.github.io/project-on-wheels/

Some other changes

The shops theme has received an update as well. I ~~stole~~ parsed the id-tagging-scheme-files and extracted the different shop types with their icons. As a result, the shops how have very nice individual icons.

As scrolling through a long list (~160) of options became cumbersome, I did implement a new way of presenting this view which offers a searchable list of options.

In the mean time, the educational theme has launched. This one features a list of languages that one can choose as school language. This list was queried from Wikidata and is very long (>1000 entries) - immediately breaking this new searchable options tool… It was amended to only show the official languages of the country the school is located in (again: wikidata info and a bit of magic). If the school still uses another language to teach, searching will probably reveal this information.

And, as usual, many translations came in and some miscellaneous bugfixes and question updates were added as well.

Towards unified tagging of schools

Posted by Pieter Vander Vennet on 8 June 2022 in English. Last updated on 9 June 2022.

For my work at anyways.eu, I’ve been tasked to make sure that all schools are in OSM - especially with capacity.

No better way to do this by making it easy for contributors to add the correct data… So, I wanted to create a MapComplete theme for education. Normally, I would open up the wiki to see what tagging is needed, but for schools there is very little tagging available at the moment, which is a mess.

As it turns out, schools are diverse and this is reflected in the tagging.

This diary entry serves two goals:

  1. I want to organize my thoughts on how a tagging model could look like
  2. It is meant to stir up some discussion.

Hopefully, some tagging proposals will come forward from from this post.

So, what is a school (or educational institute) anyway?

This is already a hard question. The openstreetmap-wiki on ‘education features’ states:

Education features are map objects and object features which relate to educational activities

Well, thanks, captain obvious.

Let’s turn to the International Standard Classification of Education (from Unesco) instead:

As national education systems vary in terms of structure and curricular content, it can be difficult to benchmark performance over time or monitor progress.

So, in other words, it is difficult as this can be highly different amonst regions. The ISCED-document however does a good job to draw some lines and to give some definitions.

What does a standard school curriculum look like?

In most countries, the school trajectory for most people (according the the ISCED, page 21) looks more or less as following (but the precise ages can vary with a few years):

Before formal education starts, kids younger then about 4 or 5 go to preschool/kindergarten. This is optional in most countries, and some education takes place, often to prepare spelling and simple math. ISCED calls this level 0

Between 4-6 and around 12, kids learn to read an write, learn basic math skills and other skills. This is called primary education and corresponds with isced level 1

Between 12 and 14/15, kids get lower secondary eduction and learn more skills and competencies (isced level 2). Between 14/15 and 18; kids get higher secondary education (isced level 3). Note that the secondary levels have a split between education preparing for (a set of) trades versus a general training which prepares for tertiary education. These orientations are called vocational and general education.

At age of 18, someone who has obtained upper secondary education, could join the workforce, could follow non-tertiary education (see below) or could enroll in tertiary education.

The first cycle of tertiary education are often bachelors (often 3 years, but 2 years is pretty common too) and correspond with isced level 6, after which a master degree (often 2 years) which corresponds with isced level 7 can be obtained.

At last, a doctorate can be obtained which corresponds with isced level 8.

If, at age 18 someone does not want to enroll in tertiary education or isn’t ready yet for the labour market yet, they can also follow a post-secondary non-tertiary (ISCED level 4) education. This is an education which is not sufficiently complex to qualify as tertiary eduction and often has a vocational training, thus a training which prepares for direct labour market entry. Note that the ISCED does not state typical ages for this education form, as it is often taken by adults too.

At last, isced level 5, officially called short-cycle tertiary education provides education to prepare for following bachelor degree, e.g. if the skills obtained by a vocational secondary degree are not sufficient to enter a bachelors degree.

What if the education is non-standard?

A good tagging scheme doesn’t break under special cases. Lets have a look at some of them to test the waters.

While most of the people might follow a trajectory as outlined above, many don’t.

The ISCED-definition leaves wiggle room by more or less defining what skills one gains in a certain education level - not at what age someone typically obtains these skills. While the typical ages are stated in the ISCED, they are not the defining features.

Some examples of non-standard trajectories could be:

  • Someone who has never had the chance on learning how to read might enroll in primary education as an adult.
  • Someone with a learning disability might be obtaining the lower_secondary skill, even though people of their age age are in higher_secondary.
  • Someone in their forties might wish to reorient their career and follow a vocational course of the skill level of a vocational_upper_secondary.
  • Someone might follow a course in music, dancing, skiing or scuba diving as a hobby in an informal school during the evenings, while still working their job during the day.

This last example also touches upon specialized schools. How should these be handled? Examples of these schools are: - driving schools or flight school. - a secondary school which focuses on arts, but has enough general skills to be compatible with ISCED-level upper_secondary? And what if this school contains a college with a bachelor degree in music too, in the same buildings?

So, this implies that knowing the isced-level of a school is very useful and often does imply the age of the pupils, we still need a way to express whom is going to this school.

Who is the school for?

By default, we could assume that most schools are normal schools where pupils follow age-adequate courses.

This is not always the case. Some schools focus e.g. on secondary education for adults, other focus on people with disabilities.

To tag this, I propose to introduce a tag school:for, e.g. school:for=autism, school:for=adults, school:for=learning_disabilities, …

If this tag is missing, one can assume that the school is for normal-abled people whom follow courses typical for their age.

In some places, schools are separated by gender too. Some schools are boys/girls only, others teach both but they are separated. ~~This might fit this tag too, but not quite.~~ Update: As it turns out, this is already handled by school:gender

At last, a school might provide different levels that are often not grouped. E.g. a primary school might include a preschool. One can even conceive very special settings, e.g. a single university building which has an embedded kindergarten for the kids of the teachers!

What does a school teach?

The further in the education system, the more specialized education gets.

Where all primary education teaches more-or-less the same subjects, secondary education already starts to specialize.

And tertiary education is extremely specialized, with faculties teaching about just one field.

I propose to introduce a school:subject-tag, which indicates what subjects are taught at a school. This must be independent of the ISCED-level. For example: a school might focus on “teaching music”, which can range from evening school for adults, to a secondary school that qualifies as isced=upper_secondary to even a college in arts having doctorate students.

Giving an exhaustive list of possible values is impossible, but some common values could be:

  • arts, music, dance, painting, …
  • driving
  • flight
  • to disambiguate, a wikidata-entity could be linked

The tag school:subject would also remove the need for various extra amenities, such as amenity=dance_school, amenity=music_school, reducing complexity. Other details and assumptions (e.g. target audience and offered education level) can be clarified as explained above.

Schools which do teach skills without general education (e.g. a driving school) could thus be tagged with:

amenity=school school:subject=driving isced:2011:level=post_secondary

A college, tagged with amenity=college thus implies isced:2011:level=professional_bachelor. Whether or not a master-degree can be obtained at that college can not safely be assumed.

Schooling method

At last, there are multiple ways to teach students. Especially secondary education has a rich variety. In Flanders, we have Montessori schools, Freinet, Steiner, CLIL, … This could be worthy of a tag too, e.g. with educational_method oreducational_method:wikidata

Schooling language and taught languages

Schools might be operated in different languages - especially important in areas where multiple languages are spoken. As it turns out, school:language is already in use for this, but wasn’t documented in the wiki. Now it is!

Other tags

Of course, there are still other well-established tags important too, such as capacity, contact information, … I’m not covering them here, as these are already widely accepted.

Post-tertiary education

Some colleges offer master degrees for people who already obtained a master degree. How to tag those?

Other questions

What is a college in OSM exactly?

Conclusion

Schools are diverse in the subjects and the level of education they teach, how they teach and who they teach. This makes tagging difficult. This post describes a possible method of splitting these subjects into orthogonal tags which can be independently measured.

This blogpost attempts to give a first attempt, but of course, I’m only aware of my own environment. There must be other types of schools which I’ve never heard of before, so if you know of something that is considered an ‘educational feature’ which cannot be tagged with the tags described above, please let me know.

Hi all,

MapComplete has been (partially) translated in 21 languages by now - an amazing feat that I could never have done alone (for starters, I don’t speak 21 languages).

(For those that don’t know MapComplete - it is an easy-to-use map viewer and map editor. It shows POI on the map and when clicking something, shows the known information and asks questions about it)

Translating MapComplete started by manually editing the translation files, making a pull request, … A next step was moving over to Weblate, where there is more support for translations.

However, one has to know where to find this translations and translating can be tricky, figuring out which piece of text goes where.

That is why I launched a new feature today: toggling ‘translators mode’ (in the copyright tab) will add little buttons, taking you directly to the page to translate the string (or to fix some typos):

So: please help to translate MapComplete in your favourite language!

Some remarks:

  • You have to create an account on hosted weblate first. If you register with the same name as your OSM-username, translators mode will trigger automatically
  • Translations take a few days up to a few weeks to appear in the application
  • If the original string has a part between { and }, just copy that part (and don’t translate it). The parts between braces have special meanings, e.g. it will show data from the object itself
  • If translators mode is active, all languages will be shown in the language picker. Pick the one you want to translate into. Is your language missing? Go here and follow the steps.

And again, thanks for helping out! You all are awesome!

Last week, we launched a new OpenStreetMap-based website: “Pin je Punt”.

This application -based on mapComplete- was commissioned by the flemish touristical agency Visit Flanders. The application is a map viewer and map editor, which requests information about charging stations for ebikes, bicycle rental, benches, playgrounds and a few more POIs.

Why?

This project was started for two reasons:

  • No one in Belgium does know where all the charging stations for bikes are
  • There are five different, provincial touristical offices. They all keep their own database of POI as benches and bicycle rental. ‘Visit Flanders’ (which is responsible for the five smaller ones) wants to unify and integrate these databases. And where better to store all the data then on the biggest geodata repository?

This project builds upon MapComplete, which aims to be an easy-to-use map viewer and editor.

The launch

The project is live since the 7th of march, the (dutch only) project page can be found here

A former radio presentator made an advertisment video as well, which can be seen on youtube (dutch only). As far as I know, this is the first professionaly made commercial which asks to contribute to OpenStreetMap!

There has been quite a few edits done by now too:

The “import” tooling

As mentioned, these touristical offices already have quite a bit of data, especially about benches and picnic tables - but also a few playgrounds are known. For this project, the agency wanted to get all this data into OSM.

Importing these datasets (after some cleanup) wouldn’t be the best idea, as the data isn’t very fresh anymore. We opted for a different approach: a new note was created for every missing datapoint. These notes are loaded by MapComplete and shown when appropriate. A contributor can open MapComplete, and then confirm that a bench, picnictable, toilet or playground is effectively there. If it is, they are invited to slightly move the item to a more accurate position using the best available aerial imagery in the region. If the item is missing or duplicate, this can be marked as such and the note is closed too.

For now, this approach has been quite a success: from the 871 created notes, 87 have been handled already - some of them by long-time community members before the official launch; other by new contributors in the first days since the launch.

A big disadvantage of this approach is that we are spamming the notes. This can be frustrating for contributors who try to respond to every note, especially notes from newcomers to give a welcome feeling to them. For them, I created another MapComplete theme showing the OSM-notes. This one has advanced filtering fuctionality, to hide import notes, but also to filter notes created by someone, not created by someone, created before or after some date, ….

A second drawback is that some of these notes will be open for years. For this, we provided some tooling which shows all the import notes of a certain campaign with the option to close them all at once. When the project is ended (e.g. in one year time), we can close all remaining notes at once.

All information about the imported dataset is available on the wiki

At last: contributors with >5000 changesets can use the import helper page on mapcomplete too. If you have a dataset about POI that are supported shown by one of the official themes, you can upload it if it is thoroughly cleaned, attributes are correct OSM-tags, the data is in geojson or csv format and your local community is informed and consented to the import.

The helper page will help you review the values, check the locations, remove duplicates from earlier open notes, remove points that are close to already existing points and finally upload it.

A screencap of the upload can be seen here

Location: 1853, Flemish Brabant, Flanders, 1853, Belgium

In my little OpenStreetMap-editor translations are provided by contributors on hosted weblate, where thousands of text snippets have been translated already in the past year - which is awesome. Thank you translators!

However, the language picker was a bit dry: it used to have codes for every langauge, e.g. nl, en, ja, ‘pt_BR’, ‘zh_Hant’… Quite boring and not really user-friendly - but easy to implement.

Today, I decided to give these an overhaul. I wanted to show proper language names in them. But: in which language should we show the language overview?

Should we show the language option in the language itself? Or should the languages be shown in the current language? Showing in the current language also means that the name of every language should be translated too - a huge task… Also, translating every language has the drawback that, if a user accidentally selects a language in a foreign writing system, they’ll won’t be able to find their language in all the “gibberish”.

Best of both worlds

I decided to offer the best of both worlds: in the menu, first language name is shown as the native speaker speaks it, followed by the language name in the current language (except if both are the same)

This means that, in all circumstances, everyone can find their language.

But, where to fetch every language name in every language?

Wikidata to the rescue

Of course, the internet must have a list of languages translated in every language. But where to find it or compile it?

I decided to have a look at one of the biggest repositories of knowledge: Wikidata. They do have an entry for every language (e.g. Dutch). To fetch every modern language, we turn to the SPARQL-endpoint with the following query:

sparql SELECT ?lang ?label ?code WHERE { ?lang wdt:P31 wd:Q1288568. ?lang rdfs:label ?label. ?lang wdt:P424 ?code }

This one fetches all languages and uses the labels in every language as their translation. With a Typescript these can be downloaded and used as translation.

The messy real world

Of course, real life isn’t as easy. There are dialects, differences in notation between Weblate and Wikipedia (e.g. zh-hant vs zh-Hant). But with a few exceptions, this can be fixed too. Some pragmatism doesn’t hurt - even though it is nice that it works for most cases automatically.

The full script is available here.

The results

Location: 0.000, 0.000

MapComplete 0.14 is out

Posted by Pieter Vander Vennet on 18 January 2022 in English.

Hi all,

Just a quick heads up that MapComplete 0.14 is deployed!

Apart from lots of fixed bugs and performance improvements, a new theme showing OpenStreetMap notes has landed! Especially the filter functionality should come in handy

And of course, there are plenty of (new) themes, such as the street lighting theme (thanks Robin!) or the postboxes and -offices map (by nicolelaine).

And -by now- over 1000 individual contributors have used MapComplete.

Hi everyone,

A new version of MapComplete has been deployed. It has a lot of exciting new themes and features, which I’d like to highlight in this diary entry.

New themes

The Flemish touristical agency (Toerisme Vlaanderen) has asked me to create a tool to help them to map charging stations, restaurants, café’s, …

This has resulted in quite some fun new themes:

At last, the theme I’ve quickly made as demo with hackerspaces and makerspaces is online as well.

Other improvements

At the same time, I’ve been doing lots of work on improving performance. A lot has changed under the hood, resulting in a way more robust and fast experience. On high zoom levels, the OSM-API is used directly, bypassing overpass all together. If overpass-api.de is down, failover to other instances is now builtin. At last, caching is a bit more aggressive and if the data on your machine is recent, no new data is loaded at all.

Earlier releases had some important features to, which I’d like to highlight as they can be important for theme builders:

  • A layer can configure filters
  • A layer can demand to have ‘precise input’, showing an extra draggable map (with e.g. aerial photography) when a new point is added to the map.
  • A new point can be snapped to an already existing road, e.g. when adding a bollard
  • A layer can define that a point can be (soft) deleted
  • The new Mapillary-API has been integrated
  • Some themes have a download-button enabled to download as GeoJson or as CSV

I need your help!

As usual, your contribution is invaluable as well! An easy way to contribute to MapComplete is by translating parts to your language. This can easily be done on Weblate and takes only a few minutes to get started.

If you want to make your own map theme on MapComplete, that is possible to. If you know OSM well, you can set up your own theme pretty quickly too. All documentation can be found here. If you have a cool theme, I’m very happy to merge them!

At last, I’m working on a small theme that shows what something is named after based on wikidata. However, I’m still searching for a decent icon for this.

A bit over a year ago, I received a phone call: We would like to have some project about nature and forests, where people can like go out and add data about those. Oh, it has to be really simple to use.

Some months and some funding later, the first version of MapComplete was born. While it had a specific focus then, it has grown a lot. Today, MapComplete is a map viewer for a certain theme, where the configuration for that theme decides what features are visible with which icons. When a map feature is clicked, known information is shown and the user is invited to add more information and to add pictures.

By now, there are about 20 themes for multiple topics, such as bicycle shops and pumps, playgrounds, public bookcases and much more. And it is relatively easy to create and load your own theme.

The ease of use and flexibility has lead to a steady increase of contributors that way. Thursday 15th of april was the day the editor saw its 500’th contributor:

Cumulative contributors

These 500 contributors made over 3000 changesets. The most popular (official) themes (in 2021) have been cyclofix (with bicycle pumps, cycle repair stations and more), trees, benches, bookcases, surveillance cameras (that one had an uptick after a documentary about them in Belgium), drinking water, artworks and defibrillators.

Theme breakdown

Want to see more statistics and fancy graphs? Regularly updated stats can be found here

Give it a try yourself on mapcomplete.osm.be, embed it in your site or make your own theme. If you have questions, get in touch via Telegram where the dev and some community members are.

And stay tuned, some more themes are in the pipeline, made by community members too.

Hello everyone,

As a coach for the Open Summer of Code in Belgium, I thought the community would be interested in reading a little about it and get inspired - along with reading how many of the teams used a map and specifically OSM.

What is the Open Summer of Code

If you have never heard about the Open Summer of Code yet, it is an awesome project. Students create an open source or open data application which a real client needs. The client -often, but not always a governmental organization- sponsors the students. During the four week program, the students get relevant workshops, gain lots of experience, have lots of networking opportunities and are exposed to other Open Source projects and ideas. This all is organized by Open Knowledge Belgium, a small but amazing organisation which furthers the use of open data in Belgium.

This year 73 (!) students participated, along with 17 coaches, making this eight edition bigger then ever. Furthermore, oSoc has been more international then ever, with lots of international students and projects and having a spinoff in Spain. This first Spanish edition had 10 students whom enjoyed it a lot. In other words, don’t hesitate to contact us. Wherever you are, you are welcome as student, coach, client of partner organization or perhaps even as organizer of your own edition.

For me personally, it was the first I participated to oSoc. I had the honor of coaching two teams, quite a challenge for someone just stepping in; but whatever challenge arose, there was always some coach with the right expertise to help out. Furthermore, with Ben Abelshausen (Xivk) and Jonathan Belien (jbelien) being coaches as well, OpenStreetMap was well represented, especially when Joost Schouppe passed by as well.

Everyone from oSoc

Maps in the projects

Lots of the projects needed a map. In the spirit of the project, we encouraged everyone to use open source and open data projects. OpenStreetMap is of course a perfect fit for this. And with so much OSM experience in the room, there was always someone to get people started with Leaflet or Overpass-turbo.

And boy, lots of projects needed a map: 10 of the 17 projects. Out of those ten projects, only one did not use OpenStreetMap. Ironically, this was one of my own projects: the project focused on visualizing sea data, thus the ESRI oceanographic map was suited better.

The other nine used OpenStreetMap in some form or another. Some used OSM as a small embedded map as extra information about some building, whereas other projects had it as a more central component; thus a map visualizing lots of data.

Ben coached another two projects. The first project, bike4brussels is a route planner for cyclists in Brussels offering many handy options, such as having a more relaxed ride, the shortest ride (which often is over bicycle-unfriendly roads), a compromise between those or a route with a bias towards the cycling network.

Ben’s other project was even more OSM-centric: the Brussels governement asked us to compare their official data with OSM. They use OSM regulary for their operations, as the official data they offer has no routable graph. In other words, the OSM toolchain is better then their own internal toolchain - but they wanted to make sure that every road and all other official data is available within OSM as well. This project indicated the growing level of trust in the project and the maturity OSM has reached over the years. We’re winning!

The developers view

In other words, lots of students worked with OSM in varying degrees - for small basic needs such as a simple map up to complex tasks such as the road completion project and route planning. However, they all agreed on one thing: OSM is very dev-friendly. This clearly shows the global effort that has been put into OSM, not only into the actual data but also into the software surrounding it - for which oSoc thanks the OSM community. Without OSM, some projects would have been more expensive or would have been totally impossible!

A great experience

All in all, oSoc was a great experience that was partly possible due to the power and awesomeness of OSM, but also thanks to the bigger open data/open source movement. The project clearly shows that an open approach works, that freedom and working together pays of in the long term.

I’d thus like to thank all of OSM for their work, the entire oSoc team (including the student, coaches, clients and organizer Dries) and invite everyone all over the world to participate next year as well to an oSoc.

Location: 1000, Brussels, Brussels-Capital, Belgium