Pietervdvn's blog

Response to “Why Book Corners won’t sync contributions back to OpenStreetMap”

Written by

pietervdvn

in

Hi all,

There was a bit of a stir on Hacker News recently around bookcorners.org – yet another website to register public bookcases.

I’m a heavy OpenStreetMap-contributor and made an OpenStreetMap-editor called MapComplete, which you can use to add public bookcases into OpenStreetMap (but it contains many other thematic maps as well).

I’m writing this blogpost as a reply to the blog post from Andrea Grandi on his choice to not contribute data back and why he made this decision.

Let’s dive in!

  1. What is OpenStreetMap?

OpenStreetMap is in the first place a big database with geographical data. As long as it is verifiable in the field, you may add this data. As such, there is a wealth of data about streets, buildings, rivers, railways, bridges, … But data about smaller and more humble objects are just as welcome: toilets, defibrillators, bookcases, benches, memorials, surveillance cameras …

Everyone may make changes, there are many tools to do this. Note that all contributions are tied to a contributor account: it gives the possibility to reach out to this contributor in case of suspicious activity. Often, the world is weird and the contributor added the correct data. If not, the contributor learned something and will hopefully not make the same mistake later.

Everyone is allowed to reuse the data for free, but our ODbL-license dictates that one must give attribution and that changes to the data must be shared under the same conditions. This latter clause is important in the context of BookCorners.

  1. What is bookcorners.org

Bookcorners.org is a vibecoded website that has their own database of bookcases, initially seeded by OpenStreetMap-data. This is important, as this means that the contagious OSM license triggers.

  1. What is wrong with Bookcorners.org?

I have a few grieves with bookcorners.org, which I’ll list below.

Andrea: sorry for the harsh criticism!

Yet another database…

First of all, there are already plenty of websites and databases listing the public bookcases. A five minute search gave me:

This is one of my major grievances: yet another database, yet more fragmentation of data, more fragmentation of the people contributing useful data…

In my opinion, OSM is a natural fit for this type of data, especially because the data is carried over to many other applications (such as OsmAnd, Organic Maps, CoMaps, …)

Legal uncertainty

I’ve given bookcorners.org a spin and created an account. Granted, this was easy: I gave my email-address, username and chose a password; that’s it!

But there is an important step missing: a license agreement. Not a word on how my contributions will be used and under what license they’ll be shared. This is a legal risk.

I’m highlighting a quote from the blog post:

It would add credentials, production safeguards, audit and reconciliation code, community processes, licensing work, and a long-term support obligation.

The ‘licensing work’ is indeed completely missing, including on the website itself. This means that bookcorners.org doesn’t even have permission to republish the images that were uploaded to it! There should at least be a contributors agreement with people who create an account on the website. If a 3rd party wants to reuse those images, the situation is even more problematic. Right now, I must assume ‘all rights reserved’.

I also found the following quote baffling:

There are also important licensing questions. A user’s permission to send a library to OSM is not automatically the same as having a sufficiently clear right to release that factual information under terms compatible with OSM

First of all, that is pretty much what “a user’s permission” should mean: the permission to send it to OSM. If that is not clear, the wording of the bookcorners.org user agreement should be improved (that is: once that user agreement is written in the first place)

Second, bookcorners.org is obligated to re-release the data as ODbL. Bookcorners.org started with an export of OSM-data and builds upon this. Changes to this dataset – including new items – must be republished as ODbL.

To the credit of Andrea, the data dump is now available under ODbL. It thus seems that Andrea has navigated themselves into a legally dubious position this way: they don’t have permission to republish yet they must republish.

This lack of legal insight and rigor is also reflected in improper use of the trademarked term “Little Free Library”, which I reported here.

OpenStreetMap and imports

OpenStreetMap is a community project. Many people are involved and no, they don’t agree all the time.

There are thus many gray areas in procedures, tagging schemes and best practices.

Imports are one of them. The community got burnt a few times early due to bad imports. One of them was an early import of the TIGER-dataset which contained (low quality) data about the US-roads. The US-community is still cleaning that up. Another burn was the unauthorized import of road data in Peru which had to be reverted, resulting in years of contributions being lost.

In the end, this is about quality and responsability. The bigger the potential impact, the more rigor needed. Importing all the roads of a single country needs a lot of discussion and thorough checking.

Adding a handful of bookcases? A small discussion on our forums should be enough, especially if one follows up on feedback. I have done many imports before; they can be time consuming, but it really depends on the size of the import.

Premature optimization

I’ve analyzed the bookcorners.org database and filtered all the bookcases that had an “external_id” (and are thus included into OpenStreetMap).

I’ve found 37.

For reference: there are over 46000 bookcases in OSM right now.

It is not worth creating a full pipeline for 37 POI, nor should this be discussed as an “import”. A (or a few) OSM-contributors may add this without any review or procedure, especially if they properly review the data.

Bookcases.org claims to do “moderation”. It is unclear wether this will be done manually or by an LLM, but if it is done manually, bookcorners.org might just as well add them to OSM-directly. Even if I take a generous 10 minutes per bookcase, this means that it would have taken 6 hours in total. This is less then it would take to setup the technical side of the flow, even when vibecoding.

Building community?

When I’ve been building MapComplete, I’ve always leaned back into the OSM-community and the wider open-data community.

I can’t possibly do everything in my own. For example:

  • Images uploaded via MapComplete are stored on a Panoramax-server. The server is hosted by Thibault, who generously maintains and pays for it.
  • The Panoramax-server software is developed by CQuest and Adrien Pavie (and probably others). We’ve had countless issues for which we contacted them directly. This was also valuable for them: we discovered some genuine bugs or missing features. When we did something stupid, they learned about common mistakes to mitigate on.
  • Did I mention that Panoramax is federated and well-integrated? This means that pictures added via MapComplete will also show up in the other OSM-editors such as iD and JOSM, making them useful for other contributors beyond just ‘bookcases’
  • For reviews, I lean onto mangrove.reviews. It has no use to start yet another review project. I’m already busy with MapComplete, let alone with developing a proper review system. And that is not even mentioning moderation of this!
  • By using OSM as backend, the legal work, legal risk and moderation effort is carried by that project
  • Some thematic maps heavily use Wikipedia/Wikidata (for example the map which links a street with the person it is named after). Here, I lean both onto the entire OSM and Wikipedia communities. I present the data in a nice frontend, but the risks are carried together.
  • I reuse the plantnet API to help identifying trees in the tree-theme

Bookcorners.org feels like a little island, drawing the bridge up because “reaching out is too much work”. This makes me sad, as the value is not in the data, but in the interactions and the things we build together. This is reinventing the wheel, once more.

  1. Future steps?

So, what are the next steps?

First of all, bookcorners.org must get a proper license agreement with their contributors. Ideally, all previous contributors should be contacted and agree to a license agreement which allows to share the data back and which allows to reuse the images under a FLOSS-license, e.g. CC0, CC-BY or CC-BY-SA.

Secondary, contributors could be nudged to add information directly onto OSM, e.g. via https://mapcomplete.org/bookcases which can be embedded as iframe. This would also dramatically simplify bookcases.org, which suddenly doesn’t need a login system anymore.

So, Andrea, what do you think about this? You can find me on Mastodon – send me a message.

More posts