While the 35C3 has many, many visitors, it also offers plenty of space to seek and build little, encapsulated worlds. FOSDEM has a lot less space but also many visitors; so, a cramped weekend (physically and intellectually) ensued.

As the FOSDEM is a conference without registration, anyone can stop by to watch a presentation, talk to someone or work on a project. It has dozens of different tracks representing the diverse world of FOSS. The devrooms differ dramatically in size and are usually packed at least on some time of the day. Therefore I decided against the initial plan of prensentation hopping and stayed both days in a specific track or devroom. I opted for: Tool the Docs, Collaborative Information and Content Management Applications and the Geospatial devroom. I tried to take some notes which I will lay out in the following paragraphs.

Tool the Docs

Introduction to OpenAPI Specification

Missed that one, unfortunaley.

Building Pantheon documentation

Nicalos Massart from Pegasys presented his take on the documentation for a Etherum/blockchain project. Personally, I was put back because of the metaphorical arch of the presentation. Staged as an arctic expedition the main points to be made were: Github wiki pages, static site generator, (Google) analytics, build a new system when the old one is at its limits. One solution, that registered as problematic with me: non-technical users need to learn Git to contribute to the docs. With a little bit of fluctuation in regards to contributors there might easily arise a knowledge gap.

Multilingual Kubernetes

Some similarities to the talk before, but a - in my opinion - more suitable solution. Kubernets uses the main git repo instead of Wiki pages and Prow to set some clever permissions on who may edit which part of the repo/docs. A backend for non-technical staff has been found with Hugo. Therefore the huge community of Kubernetes is able to operate (merge, push, …) without direct oversight of the Kubernetes team. A very clever way of easing the workload of a huge, multilingual project for everyone. The workflow will automatically tag the docs with the language to allow easy filtering in the repo.

Write drunk-test automated

Sven Strack from Provonix (the company with the devroom chair) talked about testing documentation in an automated way. They use Travis CI to test the docs and Graphana for insights into usage and creation. It was not an explicitly technical talk as Sven talked mainly about best practices in a more commons sense (have a readable source, use style guides & standards, be strict and friendly, only run checks on changed files, …). Very interesting, nonetheless.

Getting closer to a software help language

A talk about the monumental Open/LibreOffice documentation efforts. With a ton of legacy docs the team has not the luxury to work with the latest tools as they have to maintain 2.4K help files (or around 500 MB) per language, serving close to 60K users per day. So,the project takes incremental steps and is more about building custom tools (Libre Office XHP Editor) to make the work more manageable. Hats off to them.

Who needs pandoc when you have Sphinx?

Stephen Finucane from Red Hat delivered this really in-depth talk about using Sphinx for document(ation) conversion. As the air was remarkebely thin in the packed room, I could not pay close attention. My main takeaway was: Sphinx has a surprisingly large amount of Readers and Writers making it comparable to pandoc. I was not totally sold, but mainly because of a perceivable learning curve whereas pandoc has virtually none. If you are using Sphinx for your project already, this may give some valueable insight to getting more out of Sphinx.

To the future with Grav CMS

Aleksei Akimov from Ayden presented a workflow for generating docs. In short: Grav CMS / IDE for non-technical / technical writers will feed a static site generator whose output will be tested with Jenkins. Again, thin air in the room … but a break came to the resuce.

Collaborative Information and Content Management Applications

A private cloud for everyone

Jos Poortvliet from the Nextcloud GmbH asked right before his talk if he should switch his talk from explaing the importance of the privacy (which the audience may already be aware of) to a talk about the 200 coolest Nextcloud apps. The audience agreed to hear the more technical 200 apps talk and Jos switched promptly, just to embark on a delightful rant about privacy. If you want to convince someone not to opt for privacy abusing software, you can safely share this talk.

Who needs to know? Private-by-design collaboration

XWiki CEO Ludovic Dubost (also devroom chair) toke over for his collegue Aaron McSween who prepared a talk about Cryptpad. He took the energy of Jos and also lobbied passionately for privacy and zero-knowledge software. I was astounded by the functionality of Cryptpad which I took for an Etherpad clone. In actuality Cryptpad is more of a zero-knowledge Google Docs - you should really check it out. As Cryptpad is an (somewhat coincidental) outcome of work on XWiki, it needs funds to pay for the two main developers. There is a crowdfunding on opencollective.com as well as a subscription model and the team strives to secure public funding from the EU.

Tiki: Easy setup of wiki-based knwoledge management system

Jean-Marc Libs is a freelance consultant for Tiki and opted to have a hands-on tutorial of how to setup Tiki. Due to his low volume voice, the packed room and the lack of a general introduction, the presentation was not my favourite. If the mic worked well, you should have a neat resource for setting up a Tiki wiki.

Displaying other application data in a wiki

Again Ludovic Dubost, this time with a product presentation of the main software XWiki - where the ‘X’ stands for extensible. And this is an understatement because the whole presentation was made with and hosted on XWiki. But this was just the tip of the iceberg as he continued to show multiple Macros, Plugins, embedded HTML/JS (e.g. for Graphviz) and APIs. XWiki is reeeeally versatile. To the dismay of his collegue (her mobile data plan was providing the internet) Ludovic was browsing API calls from NASA’s space image of the day with great eagerness, delighted by XWikis smooth operation.

LibreOffice Online - hosting your documents

Michael Meeks with a display of rhetorical energy. If you need a sales pitch to use LibreOffice Online in your organization, this may be it. I have not tried LO Online yet, but it looks quite mature.

XWiki: a collaborative apps development platform

Anca Luca from XWiki with another showcase of the extensibility of their product. She made a point for XWiki as a starting point for web apps by demonstrating the standard features (versioning, search, permissions, …). Examples have been provided: vcalc.com, beta.hls-dhs-dss.ch

memex - collaborative Web-Research

Oilver Sauter from worldbrain.io with a sales pitch for memex; as of now, a browser extension with extended bookmarking features (tags, collections, filtering, …). The actual memex principle (connecting bits of knowledge) will be realized in the next year of development when people can share their collections.

CubicWeb - a browser for the web of data

Nicolas Chauvat built a browser extension for displaying the actual data of website instead of their HTML reprentations. It connects well with projects like Wikidata and is - from my understanding - a great tool to browse microformats and website data. It should prove very useful for building webscrapers as you can directly access the data without sifting through JS/HTML code (or the web inspector).

Document Redaction with LibreOffice

Long story short: Muhammat Kara showed a feature of LibreOffice to redact documents (now in dev version available). It is quite basic as it converts the document to a metafile which is opened in Draw and can be exported as bitmap, PDF, etc. Some people may need a tool like this. Future development will tell if the team will elaborate this solution (e.g. with OCR, or without the bitmap step).

Geospatial

Gracefully chaired by Marc Vloemanns the devroom started unaware of the clip-on microphone, so do not expect (good) recordings for the first two to three talks.

Improve OSM data quality with Deep Learning

Olivier Courtin talked about his setup for analyzing aerial imagery to improve OSM. His project Robosat.pink hits its accuracy limit at some higher 80ish percent and is therefore only suitable for validation or error checking. Besides those limitations, it does a very good job at providing a small scale machine learning operation without too much investment needed. He will try to optimize the algorithms for low-resoultion imagery in the near future.

3geonames.org

Ervin Ruci does not like properitary geocoding services and thinks that the Hilbert curve is superior to many other approaches (his geocodes show a semantic likeliness when spatially close - an “anti-feature” of what3words, for example). His geocoder is very sleek, has an API (running on a very small server plan, atm) and could need translation. A very nice FOSS project adhering to Linux principles.

TTN Mapper

Sponteneous lightning talk about TTN Mapper - the developer needs support.

Latest developments in Boost geometry

Vissarion Fisikopoulos has a Ph.D. in algorithms, which you will notice in this talk. My (ignorant) takeaway: the Boost library has some powerful functions worth checking out (e.g. different strategies for distance calculation). If you already use Boost, you are likely much more smarter than me and should check out this talk (it has some nice benchmark slides).

Continous Integration to compile and test Navit

Patrick Höhn presented his approach to realize CI for Navit (using Circle CI). I was not very sold on the quality of Navit, to be honest. It seems like a good platform for customization, yet seems not to be a bit stuck. I could be very wrong, though.

Linking OSM and WikiData

Edward Betts is living the dream: creating a very useful FOSS tool by himself. It basically tries to connect OSM and WikiData via SPARQL queries. A successful connection means the ability to pull a bounty of information from WikiData to be included in OSM (e.g. name of a place in different languages). Interestingly the other way around (OSM to WikiData) is not possible due to licensing (Zero CC vs. ODbL). Great for improving OSM quality.

Qwant

An impromptu lightning talk for qwant.com/maps.

Graphhopper routing engine - whats new

A very calm Peter Karich from Graphhopper showcased the insane speed of Graphhopper as well as new features. Firstly, there is map matching, which will snap GPS (or other) tracks to the actual street network. Very useful for sharing GPS tracks. Secondly, Graphhopper has now a scripting interface where the user can specifiy certain conditions (e.g. avoid primary roads and cobblestones). This could be very useful for apps relying on Graphhopper. Lastly, he mentioned an Graphhopper Android app in the play store which is not really an app but a experiment, yet it is rudimentarly useable.

Hikar - augmented reality for walkers

Nick Whitelegg showed his Hikar app (do not use the old one in the app store) which he has developed to blend walking routes and virtual sign posts with the camera of your device. Coming about as a seasoned speaker and teacher Nick’s talk was very enjoyable. I am a bit unsure whether the use-case is strong enough for me personally to install another app, but nice to know it exists.

Hundred thousand rides a day

Ilya Zverev from Juno Lab is a blatant liar because he actually does have 50K rides per day. Nonetheless, what he does with those is amazing (and he was very open about creating the sit-bait title). The taxi drivers of his company provide him with the GPS data and he detects OSM errors with “map matching” (see Graphhopper above). If you have about 50 traces which contradict with the map matching result, you may have encountered an OSM error. And further, he rasterized the travel direction of GPS tracks and color coded the directions, making it very easy to spot one-way and two-way streets (one of the most common errors in OSM data in Manhattan). He is thinking about opening his tile server to the public, providing the OSM community (in New York) with a very recent (daily) data source.

Open Source Geolocation

The small, but very interesting story of the history of Geoclue2, told by Zeeshan Ali. I was very intrigued by Zeeshans depiction of the open source world; makes you a little bit happy inside.

Using OpenStreetMap and QGIS to build resiliency maps

Stefano Maffulli will not be one of those who need rescue after the next San Franciso earth quake. He and his wife use and improve OSM to build up resiliency in their community. He came with a clear problem (a better system for visualization and printing, light-weight, no server) and, through FOSDEM magic, got instantly connected with people who can help. A good, eloquent talk without too much technical input; instead you’ll get some (prepper-ish) wisdom.

The rest of talks were too late for my schedule, so I headed home.

Summary

FOSDEM is kind of cool, but also stressful in terms of crowd densitiy and hustling for space. I think, I will have to balance the benefits with the timely constraints next year. But: as a FOSS conference in the truest sense, the user (who does not need a ticket) is in power all the way.

Two bits for the conclusion:

  • watch the talk of Jon ‘maddog’ Hall “2019 - Fifty years of Unix and Linux advances” (as soon as it is up - check the status page)
  • read this grumpy comment on FOSDEM which will counteract any Stockholm-y syndromes (or makes you less sad for not attending)