[Transports] – Create a railway business software and federate players in Europe: OSRD and the OpenRail association – SNCF Réseau

The ORSD (Open Source Railway Designer) software and the OpenRail association

Walid: welcome to this new episode of Projets Libres!. Today, we’re going to start this new series on transport, transport and free software. After the first episode where we talked about transportation simulation systems in Canada, this time, we will talk about rail simulation. And I’m very happy to have three people with me who are going to talk to us about a subject that I follow a little, a software called OSRD (Editor’s note: Open Source Railway Designer). So, it’s three people from the SNCF. I have with me Céline Durupt, who is an Open Data expert, Yohan Durand, who is a web developer, and Loïc Hamelin, who is a project director. And with them, we’re going to talk about OSRD, we’re going to talk about free software at the SNCF and also about foundations. You’ll see, there are a lot of subjects, it’s going to be very interesting, even for people who, like me, are not geeks about railways. That’s it, you’ll see, it’s going to be very good. Well, listen, all three of you, welcome to the podcast Projets Libres!.

Loïc: Hello Walid.

Céline: Thank you. Hello everyone.

Yohan: Hello.

Presentation of the guests

Walid: We’re going to start with the ritual introduction. I’m going to start by asking you to introduce yourself, tell us a little bit about your background and also explain a little bit how you fell into free software or open data. Céline, honor is yours. Would you like to start, please?

Céline: Hello, my name is Céline Durupt. I work on projects related to open data in particular and also open source a little less, but I would like to see that a little more the case. And it’s been since I’ve been working on OSRD projects. So I have a background in geomatics at the base. So geomatics is the science that is interested in how we can manipulate, visualize and analyze geographic data with computer tools. And I started my career at the SNCF a few years ago on passenger information projects, then now for the past two years at OSRD on railway simulation projects.

And in all these projects, I worked on how we could integrate more open data and in particular geographic open data and even more so in particular database data OpenStreetMap which, for those who don’t know, is an open and collaborative geographic database, which is a bit like Wikipedia that everyone can add, modify, complete information from the database and this database is extremely complete and precise, and it is a very interesting opportunity for many uses that I could detail in the rest of our exchanges. So, I work at the SNCF to support teams that have data needs and raise their awareness about open data and help them consume open data. There you go, I’ll detail later on OSRD what it means.

And then, on the free software part, it’s always something that has attracted me, that has interested me. In geomatics, we have a very large free software called QGIS, which you may know. It’s really a field in which there is a rivalry between the free software community and a very large publisher that I won’t name that publishes competing software. This is an area in which we quickly have the opportunity to see how interesting it is to contribute to free and collaborative projects rather than consuming proprietary software. It attracted me to these subjects as soon as I was studying and I’m happy to have been able to continue them at the SNCF.

Walid: Great. So, two clarifications. The first is that I will put in the transcript a link to your conference that you gave at FOSDEM in the Transport track in 2023. Very interesting conference. And the second is that for the only ones who want to know more about OpenStreetMap, I invite you to go see the episode I did with Christian Quest not too long ago. And we’re doing an introduction to OpenStreetMap, so that’s it, it’ll be quite appropriate. Yohan, for your part, please, would you like to introduce yourself?

Yohan: Yes, so, Yohan Durand, soon to be 20 years of loyal service at the SNCF. I haven’t always been a developer, I’ve done a lot of jobs, including that of train driver. For health reasons, I had to retrain and I chose development, out of basic passion, out of passion for this world. Since I’ve always had a strong appetite for free and open source, it came quite naturally, my approach to OSRD. And by dint of discussion, I joined the project. I am a front-end developer in OSRD, I mainly deal with representation and integration. For some time now, I’ve been more specialized in graphics and end-to-end testing .

Walid: That’s it. Great, thank you. Finally, Loïc, would you like to introduce yourself?

Loïc: I’m Loïc Hamelin, I’m in charge of the program and at the very origin of the creation of the project, 2000 years ago. I started at the SNCF 26 years ago now, as a train driver, like Yohan. Then, I was lucky enough to be able to go back to school and I studied, in particular in computer science. This allowed me to see the challenges of information systems and software development. With operational experience, I quickly realized that we didn’t necessarily have all the tools necessary to help with operational management in the event of a disrupted situation and to help with decision-making to regulate traffic. Thanks to other professional experiences, in particular subsidiaries at Systra, where I did operations studies and I used simulation software, I saw that there was potential for this software that could be developed in-house and then used to help with operational decision-making, to help resolve disrupted situations.

And then, why OSRD in open source? At first, it was something I was told to do. It was my manager at the time who told me “ok, you want to make a simulator, you do it in open source”. And it stopped there. And in fact, now it’s in the project’s genes. I don’t see how OSRD couldn’t not be open source, because it makes so much sense in our case, to be transparent about simulations to explain why we make these investment choices, for example, because that’s what simulations are for. This often leads to heavy investment choices on the network, or rolling stock, or many things. But since there is a lot of public money at stake, the tools that led to the decision must be as clear as the result.

Loïc Hamelin

And so, it’s intrinsically, in a single word, OSRD must be open source because it must be auditable, since we must have confidence. And since it is a tool that must also be common to help share results, but also to help the rail system as a whole to have a common basis to avoid the problems of interoperability of information systems, the problems of data models that are completely different and above all that they are capable for us, for the railways, to simulate on a European scale to make the train competitive with road or plane.

Walid: What I didn’t say earlier, I talked about SNCF, so you work for SNCF Réseau, so the infrastructure manager. (Editor’s note: the SNCF group’s organizational chart)

Loïc: SNCF Réseau, the infrastructure manager, yes. We are in a department called the General Directorate of Operations and in an entity that deals with the digitalization of the operations professions. So, precisely, to provide operators or research managers with modern digital tools to help improve the rail sector or promote the railways.

What is OSRD?

Walid: Very good. The first thing we can do now is to start by defining for listeners who are not familiar with this field, what is OSRD in fact? What is it and what is it for?

Yohan: What is OSRD? OSRD, I like to say, when I talk about it, that if I find myself in a country where there is nothing, there is not a rail, there is not a train, there is only slab, we can decide to trace, via our editor, the system, add signals, create trains, run them on this system. Create last-minute train paths , at the last moment or according to a temporality, decide that between two trains scheduled in a timetable, we will be able to balance a train in the middle, make sure that it matches, that it passes, that it stops at the right place, that it arrives at its destination within an acceptable time scale.

As a result, OSRD is a complete railway design simulator, as its name suggests, where we will be able to manage our infrastructure finely, in detail, create the timetable on the timeframes we want, in the places we want, create the site at the last minute and edit all our trains, all our rolling stock. All this based on open source tools, so OpenStreetMap, as we talk about it, for visualization. We were on D3JS until now, so an open source library to draw graphs.

Yohan Durand

Now, we use our own methods. Does anyone have anything to add?

Loïc: I’ll complete it. Basically, OSRD’s ambition is to really be the digital twin of railway operations, both in the theoretical part, i.e. before trains run, but also one day, in the longer term, by being this digital twin that helps operators make decisions.

And in particular, as Yohan said, there is a part that is very close to the operational, which is to find a path in space and time for trains that would need to run a week before the day of operation. Rail is not like the car, you don’t decide that you’re going to drive in an hour, it’s not possible. Everything is planned years in advance. So for last-minute train paths, for trains that will run at the last minute, we will need specific tools to do that. And currently, it’s more done by hand. It is by business experts. And we need to be even better to promote rail freight, once again, vis-à-vis the road.

Loïc Hamelin

The license used in the OSRD project

Walid: Before we get into the genesis of OSRD, right after, what is the license or licenses that are used, that you used for the OSRD code?

Loïc: So we’re on LGPL V3. We chose this license in part because we want to be able to keep proprietary parts. If we have intelligent algorithms that we don’t necessarily want to share, we want to be able to keep them proprietary. That’s a first reason. The second reason is also that there are software publishers and we don’t want to kill them. We just want to agree on a common base and for them to do business with the parts they have developed themselves. If they do smart stuff, there’s no problem for them to keep it to themselves and do business with it.

But the common base that allows everyone to exchange, to agree on the calculation results, really on the basis for simulating the rail system, we really want it to be something universal, but really shared at least on a European scale.

Loïc Hamelin

Walid: The LGPL allows you to embed free code in proprietary tools.

Loïc: Yes, but on the other hand, if someone from the outside who has developed his own modules that keep proprietary, who wants to modify the core or correct the core of the free part, he is obliged to share it with everyone. So, normally, it’s a win-win.

The languages used in OSRD

Walid: Absolutely. What languages do you use inside OSRD?

Yohan: If we start from the back-end, we will have a part in Rust for endpoint management, and a part in Kotlin for what is the core of the calculation. We’re going to have a little bit of Python for our test batteries, especially for data management. So on the front end, we’ll have React, TypeScript, with a whole bunch of libraries added to it.

Walid: ok, so it’s Rust, there’s no C / C++ what?

Yohan: No, in this case the choice was made on Rust.

The genesis of OSRD

Walid: ok, let’s talk a little bit about the genesis of OSRD. I would like to understand the path that led to the creation of OSRD. That is to say, at the time it was created, what is the panorama of the different solutions? What makes you say to yourself “Ok, we’re going to create our own free product to do this”?

Loïc: The decision was really taken in 2019 by the SNCF Réseau Executive Committee, to say: we are going to develop our own solution.

We were and still are dependent on products on the market that do not communicate with each other, that act like black boxes, that function like black boxes, on which, as I said earlier, we make investment decisions that are extremely important.

Loïc Hamelin

At SNCF Réseau or in the rail sector, the heart of this software is the calculation of the rate. The rate calculation allows us, by describing the rolling stock and the infrastructure, to say “if a train leaves at 12:04 p.m. from Paris-Lyon, it will arrive at 2:09 p.m. at Lyon-Par-Dieu “. We calculate it, it’s not a coincidence, it’s all calculated. The algorithm is quite simple, it’s Newton’s physics, MA equals the sum of forces, and then we do numerical integration. It’s a little more complicated than that to implement. There is no industrial secret behind this thing. And we have a lot of this algorithm for calculating the rate, several, let’s say, at SNCF Réseau or SNCF. And we have suppliers who use other software, who use their own algorithm for calculating the market, even though it’s ultra-strategic. Because from the travel time, we will be able to know if we can add more trains to a line, therefore more capacity consumed or more passengers transported or goods transported. So having several black boxes that we don’t explain and that give different results was not acceptable. So, the decision was to say: we had a little simulation experience, we had made prototypes, we knew it was feasible. And so, when we presented this to the Network Executive Committee in May 2019, the decision was among others, but it was to say: “ok, we’re launching the development of our own simulation tool”. It took time to implement, because it’s something that requires development skills and budgets.

Walid: When you say “we”, who are you?

Loïc: We are the people I was with at the time. There were two of us at the very beginning. There was me who had proposed, there was a second person that we hired for a physicist, Juliana, not to mention her, who came with me to the team to do the first lines of code in 2019. And then after that, we stayed two years as a very small group and we had already released code, which was what was, but we weren’t a software publisher to do that either.

We were lucky enough to benefit from European grants, which allowed us to structure a team, really, to recruit a lot, and to really develop the OSRD solution as it is now. We started it roughly three years ago, really the big development. We first went from 4 to 8 and then now we are 50 basically. So, we develop the OSRD solution internally, so when I say internally, it’s an integrated team in the offices of SNCF Réseau. There are people from the SNCF group, there are people from various services, there are freelancers, there are people in umbrella companies. There are a lot of profiles who are with us, but who are integrated as a team of about fifty people.

Access to algorthimes and data

Walid: It’s something that comes up often in the first episode I did on the subject of transportation with the Transition team, so they’re from Québec, they work at Polytechnique Montréal. Basically, they said more or less the same thing: we take software, they do calculations, we never manage to find the same result and we have to incur expenses on it. They also said another thing, which is: we train people on proprietary software when we could train them on software that we master and that we are able to explain. It’s a fairly recurring subject, access to algorithms and access to data.

Loïc: That’s clearly what led us, in addition to the fact that we want to favor the train, in addition to the fact that we want explainability, of all that. It’s really about understanding what’s going on. The first need is to understand what is happening in software. And so, what could be better than having access to the code, or even having developed the code.

Walid: And here, OSRD, is used mainly by SNCF Réseau or also the other entities of the group use it for certain specific needs?

Loïc: So, on the last minute furrow part, we opened it today. Really, it’s not a joke. We have opened it today to the first freight railway companies for a test period because it is the first launch. We’ve also been using it for simulations for a little while, but more with beta testers, let’s say. We’re not finished, even though we’ve been working on it for three years. It’s a software that is so complex, that requires a lot of development and development takes time… and which is very ambitious, once again. So, we already have the first real use cases at SNCF Réseau. So, there will be customer users, especially such as railway companies (RU), from today. And it is possible that another, it will be a European railway company, will use it. They will do the first tests from… We’ll see them next week too. It is possible that it will start to spread in Europe. And then, we also collaborate with the Swiss. I don’t know if you had the information. I think we’ve talked about it before.

Walid: Yes, you talked about it in the conference you gave at Open Source Experience on business foundations with Amel Charleux.

Loïc: Amel, yes. So, frankly, they’re not going to use our solution, but there’s a collaboration with the Swiss railways that’s very interesting, that has brought value to us in the software. We make developments for them. So, normally, everyone wins. So, that’s another form of using the OSRD or the OSRD team or the development capability of OSRD.

Walid: Before we move on to the next part, when you go before the SNCF Réseau executive committee and pitch this project, from the start, there is the idea of saying: we are making a common European platform. Is it something that was there from the start or is it an ambition that revealed itself later?

Loïc: No, what I proposed to the Executive Committee was really to have our own simulation tool. After that, there is an infinite number of ways to develop it. We could have made a thick client, we made a web client. We could have done something focused on France. We could have made a proprietary tool. All that, afterwards, are choices that come later.

I still had the long-term ambition to open it up as much as possible. We also needed a neutral space to put this tool. If we want it to be reused by other railway companies or other infrastructure managers, it should not be a purely SNCF tool. It had to be put in a neutral space and there should be shared governance or clear governance on this software as well. And that’s how we end up with Open Rail afterwards.

Loïc Hamelin

Necessary, available data and publication obligations

Walid: Yes, we’ll talk about it after the Open Rail foundation. The next topic I would like to touch on is the topic of data. Because for a simulator, you need data. In fact, what I would like to understand is when you start the project, what data you have, what data sources, and therefore how you use it and how you also contribute to the enrichment of this data, both public data or your internal repositories. In fact, basically, how is it organized, Céline?

Celine: So, in order to work, OSRD needs three different types of data. First, data that makes it possible to describe the infrastructure. The infrastructure of the rail network will be where the tracks are, so the rails, where the railway signals are, where the stations are, where are all the elements that we need to know to know the origin and destination of the trains that we are going to want to trace, but also all the elements that will impact the calculation of the rate, that Loïc was talking about earlier, so the gradients, the signaling systems, the speed limits, etc. All of these elements. I will first present the three types of data, and then I will explain the different sources that we have. The first type of data is infrastructure data. The second type of data is rolling stock data. Here, we will describe all the equipment that can circulate on the tracks, so locomotives, self-propelled vehicles and the wagons that can be attached to them. For this data, we will have physical modelling information, such as their weight, length, traction and braking power, and their compatibility with the infrastructure as well. For example, what signaling system they support, what is their gauge, is this gauge compatible with this type of infrastructure. And finally, the third type of data that we need for OSRD is the timetable data, so to know what are all the other trains that are already scheduled on the network, for the functionality that allows you to add a train, to be able to know which trains are already scheduled and how the train you want to add will not disrupt all these trains that are already scheduled. And so this timetable data will be composed of a train ID and then all the schedules at each of the crossing points of this train.

So three types of data: infrastructure, rolling stock and timetable. For these three types of data, currently for SNCF Réseau’s internal uses, internal repositories are consumed. I’m not going to go into the details of these standards, I don’t think it will interest listeners. But we also looked at how we could access this data with open data, for several use cases, in particular that of having a public version of OSRD that can be tested by people who are interested, and also that of being able to collaborate with people outside SNCF Réseau, whether they are railway companies, research laboratories or companies that would like to develop functionalities to complement OSRD, as Loïc mentioned earlier, or for any other use case that could be imagined. As a result, to meet these open data needs, we will not have the same answers depending on the type of data.

For infrastructure data, we have a first option which is to use OpenStreetMap data. This is the presentation I made at FOSDEM two years ago. The limitation is that OpenStreetMap data is quite different depending on the country. So, we made a first version that works quite well. It is not 100% complete, but it is functional in Germany, which is the country where we found the most complete OpenStreetMap data. Unfortunately for France, the OpenStreetMap data is not complete at the moment, especially because we lack information on signage.

Céline Durupt

So, we thought it could be interesting to complete this data with data from the SNCF Open Data, since there are quite a few datasets that are open data on the SNCF website that you can consult. But unfortunately, there is not all the data we would need for OSRD which is 100% open data. So, we are taking internal steps to try to publish these datasets so that we can complete them. But these are dynamics that are quite complex and quite slow. So, I hope that we will get there one day, but it’s difficult to put a publication date on this data at the moment. It really depends a lot on the internal political dynamics in relation to these subjects. Sometimes, we are in rather open phases where we will publish a lot of new data. Then, sometimes, there will be phases of hindsight where we will say, in the end, it’s not good to do open data. We will stop publishing new datasets. Then maybe it goes the other way, it’s quite difficult to understand exactly why and how there are these waves. But do you want to add to that Loïc?

Loïc: We also have obligations to publish datasets at the European level in particular. It’s even less complete than what Céline describes, but I’m hopeful that in the future we’ll have enough data on a European scale to allow us to connect OSRD to a European infrastructure.

Walid: Are they European bonds?

Loïc: Yes, it’s called the infrastructure repository, the RINF (Editor’s note: Registers of Infrastructure). Little by little, Europe is forcing infrastructure managers to publish data. It’s going gradually. First, the diagram of the lines, then the remarkable points on the lines, and then afterwards, we will make the tracks, then afterwards, many other data to achieve a functional infrastructure on a European scale.

Walid: It’s a very long-term project, on a European scale.

Loïc: Yes, on a European scale, it’s a very long-term project. It is a negotiation process, I imagine, between the infrastructure managers and Europe. We don’t all have data of the same level of quality, so inevitably, some are more advanced than others. So it’s harmonisation at the European level that is quite complex.

Céline: To summarize infrastructure data, for the moment, it’s complicated, but OpenStreetMap is super rich, it already works in some countries and we are hopeful that we will be able to continue to enrich it in the years to come, either via OpenStreetMap, or via an open data policy of the SNCF group, or via European legal constraints that Loïc cited. Then, on the timetable data, there is already a part of the timetable data that is in open data, the same, made available by SNCF. Unfortunately, it does not contain all the trains, in particular it does not contain all the freight trains that are not public. But well, I admit that I don’t know at all if it is intended to be expanded in the years to come, but there is already a very good basis for work. And finally, on the last data we need, so the rolling stock data, this is data that is quite sensitive since it belongs to the railway companies and equipment manufacturers, so we can’t publish it as it is. But on the other hand, we have a project that should be completed very soon, at least at the beginning of 2025, which is to publish dummy rolling stock, but which has consistent physical characteristics. It’s not a TGV in particular, but it’s going to be a high-speed train that could be any of the TGVs or other high-speed trains from another country and that could be used to trace high-speed trains on OSRD. And thus to have an open data model of each of the main types of rolling stock.

Walid: And you, you collaborate, do you contribute in any way as SNCF Réseau on OpenStreetMap? You also have contacts with other entities that could, I think in particular, there was a State of the Map conference this year of people from SNCF Voyageurs who explained how they were also contributing, I think, on station data (Editor’s note: in fact it’s AREP, see OSerM: mapping stations today to improve them tomorrow, in French). I wanted to know if you had contacts or if you were dependent on their data too in some way.

Céline: So yes, there are several SNCF services in the broad sense that are interested in OpenStreetMap data, that consume it and enrich it for a lot of different use cases. I don’t want to name them all, but we are in contact internally to share on these subjects. It doesn’t always overlap exactly. For example, the station data that is collected in OpenStreetMap by Transilien, we don’t need it for OSRD, because what interests us is rather the description of the railway network and not specifically the stations. But on the other hand, it’s true that there is a common interest in this data. And it’s true that we have our small internal SNCF OpenStreetMap community to collaborate on these subjects, to give each other information on how to consume data, on the companies with which we can also work to support us in these subjects. I’m taking the opportunity to advertise for the french Federation of OSM Pros. If you are interested in this subject, you can check out their website. It is a federation of companies that know OSM very well and can support projects either in data collection or in the use of open data. There you go.

Walid: I’m going to put a link to a recent conference (Editor’s note: see this presentation conference of FPOSM during the AlpOSS conference, in French). I don’t remember in which room, precisely, he presents the Federation of OSM Pros. It’s a very interesting subject too. It’s really an issue that comes up in many different areas and to have a common and to be able to enrich it and take care of it and make sure that it contributes to everyone. It’s also interesting because in the episode on OpenStreetMap, Christian Quest, he says that indeed, the number one country for mapping in OpenStreetMap is Germany. So indeed, it confirms well.

Walid: And if I can add something, not specifically about… Well, in fact, it also has to do with open data, but on data in general.

We have a very big challenge on the quality of the data. And this is something that is very difficult to evaluate and evolve. As a result, I think that open data is an opportunity to pool our efforts to improve the quality of data and to allow a critical evaluation by any reuser of this data. In this respect, this is one of the great strengths of open and collaborative data, which makes it possible to achieve better quality than internal data, which necessarily has fewer reusers and therefore fewer people who will be able to alert if there are quality problems.

Céline Durupt

Loïc: And if we have an open dataset, it allows us to reach more people too. That is, there will be more OSRD users with an open dataset? And Céline spoke earlier, there are people who do very good things in research for the railways, in labs, and who don’t have access to software or infrastructure description data. And if we manage to provide them with software, algorithms, and datasets, they will be able to contribute to the improvement of the rail system. And it’s super interesting because they work with their means in their labs and unfortunately, we can’t capture the value they produce, when the entire rail system should benefit from it.

Walid: Have you already presented your work in some labs?

Loïc: yes, we’re working… We have contacts with two labs in particular. It was Gustave Eiffel in Lille, Gustave Eiffel University, and then with UTC in Compiègne. We also exchange with the Ecole Centrale. The rail research community is quite active, so we often have interactions.

The OSRD Team

Walid: Next topic, I would like to talk about the team. What are the different profiles in the team? How they come to work on this tool. There, you said earlier, Loïc, that there were about thirty people working on it. So, what profile will we find in the team?

Loïc: There are 50 of us.

Walid: ah, 50, sorry.

Loïc: 51, 2, 3, 4. Honestly, I don’t know, I’m sorry. The profiles are diverse and varied. And I think that’s what makes this team so rich, it’s to have these diverse and varied profiles. That is to say that we don’t only have people from computer science schools. We clearly have very brilliant computer engineers that we have recruited, either directly within SNCF, or through other contractual arrangements, let’s say. Yohan talked about it, maybe he will explain the UptoDev program of e.SNCF which has allowed a certain number of SNCF staff to retrain in IT development. I don’t know if we still have self-taught people, but we have retrained people, but not by the SNCF. We have physicists, physicists for that matter. We have a little bit of all profiles. On the UpToDev side, if you want to introduce the people of the team, Yohan.

Yohan: Yes, so UpToDev was an internal program of the SNCF, at least of the SNCF Groupe, on the e.SNCF side essentially.

The idea was to train a few hundred new developers for profiles that were essentially in retraining. Double benefit for this, the first was to have a workforce quite quickly and in addition to having a workforce that already has potential business knowledge, and that can bring this knowledge in future developments. So we operate a little bit like a control room, that is to say that we will be sent on missions where we are needed, within the SNCF companies.

Yohan Durand

Really there, there are also a lot of profiles. I did sales work at the very beginning and then I was in attraction as a train producer. Yesterday, I ended up in the offices in the IT department. Besides that, I have a colleague who was also a train driver, and another who was in sales. We have a former horarist (Editor’s note: to find out a little more about the schedules, see this conference from the Espace des Sciences in Rennes, in French). Yes, it’s quite varied in itself. So we actually did a bootcamp over seven months, with a title passage at the end. So it’s clear that Loïc was talking about the engineers who arrived in the project first. I’m not going to lie, when we arrived, he made us dizzy a lot. They always make us dizzy, but in the end, in fact, it’s enough benevolence in the project for us to move forward at our own pace. In any case, for DEDs, that’s how it is.

Walid: Two questions on that. The first is, is it a personal motivation to come and work on this project or was it Loïc who found and recruited you? And the second is, what does it feel like as a former train driver to work on a tool? So, on the other side of the fence, were you aware of all the implications that this represented?

Yohan: Well, not at all. Besides, in addition, I can really talk about it since I am a scammer. I literally invited myself to the discovery day to be hired. No, in fact, the idea was at first, at the end of the training, to have an internship. As a result, we had internship offers everywhere, including OSRD, which wanted to promote retraining as well. And when we arrived in OSRD, all we knew in the end was that it was a big project. A big project at SNCF Réseau, and which was potentially looking for profiles from traction, so from the driving profession in particular, or others, and above all profiles rather full code, not low code as we can also have on the DED side. And indeed, when we arrived at this day of discovery and we already saw the scope of the project, having a little business knowledge, the gains it could bring, and the challenge too, we were quite seduced directly. But before arriving, we still didn’t know exactly where to set foot.

Walid : And for you, Céline, how did it go?

Céline: So, for me, I had finished my previous experience at Transilien, where I had worked on the data collection of an OpenStreetMap point for train stations. And I was looking for another project to continue working. I wasn’t necessarily attached to staying at the SNCF, but I wanted to work in a project that made sense, so in a company that makes sense environmentally and socially. So, it was also fine for me to stay at the SNCF. And it was also important for me to work in the subject of open data and the icing on the cake in open source since I had tried unsuccessfully to open source the software I was working on at Transilien but hey, that’s all another story. So I met Loïc a bit by chance, I don’t know if it wasn’t Simon (Editor’s note: Simon Clavier) who put us in touch

Loïc : I don’t know if it’s Simon or Bertrand but one of them told me “there’s Céline who is changing jobs, you have to meet her, she’s so good”.

Céline : (laughs) and so, I met Loïc, and he told me “we have problems with our data. Do you want to come and stick your nose in there and see if we could do more open data”? I didn’t know anything about railway operations, but I had been working at the SNCF for several years. So, I said to myself “well, it’s okay, I worked in stations, trains, it must be the same”. And in fact, to my surprise, I still had a lot to learn. And I did learn a lot of things, a lot of new concepts. in this project. And then, I hope that we will do more and more open data. In any case, we are working on it.

Loïc: And when I say it was Simon Clavier or Bertrand Billoud who introduced me to Céline, that’s also how it happened for a number of us. That is to say, it’s a bit of word of mouth, a bit of the networks that are set in motion. Hey, I’ve heard about this project, it’s interesting. We have pretty much siphoned off the EPITA classes that come out of engineers who are very solid, computer engineers. At the beginning, and then afterwards, through meetings or things like that, we did a bit of lacework, we recruited one by one. We brought in potential candidates, they spent a day with us. We didn’t do one-hour interviews as is usually done. In fact, the candidates spent a day, they discovered the team, they discovered the project, we discovered each other together. And in the end, either we decided to continue together, or we stopped there. But we have also, in the recruitment method, we have been a little innovative in this way. Because I have the impression that we can’t realize on both sides during a one-hour interview what the person is, what the project is, what the atmosphere in which I’m going to work is, the stakes, all that, the technicality or not. A lot of word of mouth, network operation, then a slightly innovative method, well innovative that’s a big word, but the fact that just spending a day together, I think it’s really good. It made it possible to catch candidates that we wouldn’t necessarily have had, with a one-hour interview, because it’s too formal. In fact, what is needed is more for us to seduce the developers than the developer to seduce us.

So I prefer to be transparent about what we’re going to offer and we try to be as transparent as possible. And usually, people like it. What Céline said too, obviously, works in our favor. We have a project that makes sense in terms of transport. I don’t want to greenwash, but there is bound to be this idea behind it. We do open source. It motivates developers even more. So we are quite attractive. I dare to hope so. And we have a project that, technically, is quite interesting too, because we take up technical challenges that are not very easy. And then afterwards, we make the networks work. And then, we are forced to say, the classic recruitment methods by call for tenders, we are obliged to complete the team like this.

Loïc Hamelin

Céline: If I can just add something to everything Loïc has just said, I think we can still say that we have a turnover rate that is particularly low, especially for an IT project. So here, I think it’s also a validation of the interest that the people in the team have in the project.

Walid: Before we move on to the financing part, my last question is, is OSRD, now, a project that is known or recognized within SNCF Réseau?

Loïc: I think it’s starting to be known. Granted, I wouldn’t have that pretension. We have to prove ourselves again. I’m confident, it will come, but we still have to prove ourselves, that’s clear. The production launch, once again, is today, you see. So on the studies, the first users, they seem seduced. They know that there is a huge potential. We have to move from potential to reality, to concrete. But we’re starting to be known and I think the first results are encouraging. So I dare to hope that it goes for more.

OSRD funding

Walid: That brings me to the next question. We talked about the genesis of the first financing, European funding, etc. What I would like to understand is how do you finance OSRD? Is OSRD currently fully financed by SNCF Réseau? There have been European funds. Do you have other funds on the side, other additional financial resources? How does it work?

Loïc: Réseau contributes to the financing of OSRD, that’s clear. We applied for a European funding program called CEF in 2021, Connecting Europe Facility, the program that provided us with a large part of the funding. But not only that, European funding must be backed by a contribution that has been given by the Ministry of Transport, basically. So we also have funding from the State directly for the project, so the State, Europe and SNCF Réseau. So that allows us, yes now, to be about fifty people. It’s a fairly substantial budget. We are back to a round of funding under the CEF 24. So again on the same principle, the same financial package.

Walid: long cycles?

Loïc: three-year cycles.

Walid: ok, it allows you to see.

Loïc: It allows you to see. In three years, frankly, we have developed a lot. I can’t wait to be in three years to see where we’ll be in three years. It’s quite long cycles, but it allows us to structure ourselves, to do coherent things and to do things over a fairly long term.

The governance of OSRD and the creation of the OpenRail association

Walid: I would like to talk about governance. There, it will also become very interesting. OSRD’s governance. At the very beginning, we mentioned the OpenRail foundation. We come back to it a little to understand the path that comes to make a foundation. You start from the idea of saying we’re going to make a tool, it’s going to be open source, to… A lot of things happen and we arrive at the creation of a European foundation. Can you explain to me a little in general the thought, the path that has come up to this point?

Loïc: In fact, the path is simple. It’s OK, like I said, I was told the project has to be open source. OK, I’m making it open source. What is open source for? I knew open source, but no more than that. I used open source software and all that. I’m not a free software player from the very beginning. And I unrolled the open source model a little bit. What is open source? How does it work? What’s the point? Etc. And I came to the conclusion that we needed this neutral space. If I wanted to have external contributors, because that was also the goal of doing it in open source, I needed a neutral space. So I started with the idea of creating a foundation around OSRD. Then I took on an intern at the time, who was helping me find a bit of the open source business model, and who came across a thesis by a researcher from Montpellier that you know, whose name is Amel. So in 2020, I think, we met, we contacted Amel Charleux from the University of Montpellier, who did her thesis on open source business models. After several exchanges, she tells us, maybe a foundation specific to OSRD is not enough. Have you looked at business foundations, things like that? At the same time, I was in discussion with Simon Clavier, we were going around the existing foundations, Eclipse, Linux, Apache, and I didn’t know what to do. I could see the interest in computer components, in specific things, but I come from the business and I didn’t find myself in traditional foundations. So I pushed this idea a little bit to develop a specific foundation around the railways.

In the discussion group with Simon, there were already the Germans and the Swiss, who were there. And I had contacts with the International Union of Railways, where I said to myself, there are 200 UIC members, who are infrastructure managers or FUs (Editor’s note: railway companies) around the world, who represent a potential of interesting contributors to a railway business foundation. And I’m sure they might be interested. So, we put SNCF, DB, SBB and the UIC around the table. Amel helped us draw the first statutes, let’s say. And then, it took us three years to really get the foundation out of the way. It was an iterative process. We obviously worked remotely, a multicultural environment, even if it’s Europe, it’s not necessarily easy. But we are releasing the royal decree in January 2024. It is an international non-profit association, so Belgian AISBL , which officially creates the OpenRail association, because between the moment we start filing the statutes and the moment they have accepted, Belgian law changes and we no longer have the right to call ourselves a foundation, so we call ourselves OpenRail association.

Walid: We’ve already talked about the AISBL several times, including the last time on the episode with Gaël Blondelle on Eclipse, where he talks about what an AISBL is and why they chose it, etc. It’s quite interesting to dig into the … Why in Belgium? We also talk about it in the episode on the Tryton ERP, also about why Belgium and what the statutes are and what it’s for. It’s quite interesting.

Loïc: We’re going to give credit where credit is due. Gaël had introduced us to Eclipse and the AISBL and that’s what inspired us a little in the choice.

Walid: What’s interesting about all this is that you say, the idea of creating it in Belgium is to have a more neutral ground. Because if it were in France, potentially, it would be associated with SNCF.

Loïc: That’s right. And in fact, I think that at the time, in any case, in France, you couldn’t create a foundation. A foundation is very specific in France. You have to pay, I don’t know how much, 150,000 euros to launch the thing. It’s really very structured. And what is done in open source was the AISBL at the time. I didn’t have any preconceived ideas about it. The UIC had an office in Belgium, there were some facilities to communicate with the Belgian authorities. So that’s the choice that was made. Once again, it is also because Gaël had told us that Eclipse Europe was a non-profit organization.

Walid: One of my questions was precisely to know, was the choice of Belgium a choice to get closer to the European institutions?

Loïc: No, that’s not it. That’s not what motivated the thing. It’s not bad, but that’s not what motivated the thing.

Walid : Is this work of creating the foundation time dedicated to working on this time during your SNCF Réseau work, or is it something you did on the side? Finally, how did this long time working on the subject go?

Loïc: So I don’t work 35 hours so it wasn’t in the 35 hours. That’s it personal investment. I speak for myself, but it’s the same for my European colleagues, or UIC. It’s something that we have taken on our own, at our own expense. At first, it’s not something where I’m going to ask permission to work two hours a week on it. It’s done a bit under the table. On the other hand, I kept informed of the progress of the creation of the association. For me, it was clear, but I didn’t ask for extra time to work on it.

The governance of the OpenRail association

Walid: And in this foundation, this association, how does the governance work? Who is represented? Are they natural persons, legal persons? How does it work?

Loïc: So, we have three levels, or four, I don’t know, of membership members. So, we have Platinum, Gold, Silver and associate members. Four, that’s it. The platinum members are the ones who created the association. There are seven seats and there are four that were at the origin of the creation. So, they took four of the seven platinum member seats in the association. For each seat, there is one representative from each group. That is to say that for us, SNCF, SNCF group, in fact, it’s not Network, it’s SNCF group. Frédéric Novello represents the SNCF group and sits on the board of directors of the Open Rail association. We have Nicole Göbel, who is the CEO of DB Systel, the equivalent of the CIO of the Deutsche Bahn group. We have Jochen Decker, who is the CIO of SBB, the Swiss railways. And we have the financial director, in particular, but not only, the director of Jean-Michel Evanghelou, so of the UIC, who represents the UIC association on the board of directors. After that, we have other representatives, at Infrabel, we have a deputy CEO. We have had the IT department for three weeks, we have the IT department of the ONCF, so the Moroccan railways. That’s great news. It was almost my Christmas present, that OpenRail was coming out of Europe. And then there are the associate members. We also have the Flatland association . And so, each entity has a representative. But it is rather entities that are members of the board of directors. We don’t rule out the possibility of… I do not want to get ahead of myself, but it seems to me that you can be an associate member individually. Honestly, I don’t have the content of the statutes in mind anymore, especially in all the details. But it seemed to me that you could be an individual associate member.

Walid: So each member finances the running of the association through a kind of subscription.

Loïc: That’s one of the reasons why we made our association too and not use the classic Linux, Eclipse and Apache foundations. So Apache is apart because it’s very cheap. But Linux and Eclipse, to me, seemed really very expensive. The entrance fee was quite high. We wanted a lower entry ticket to get started, because we didn’t know where we were going. So, depending on the level of membership (Editor’s note: membership), the membership fee is more or less high. The highest is 25,000 euros per year. I think it’s 5,000. The lowest, associate members, it’s free.

Walid: There are no employees in this area.

Loïc: No, no, no. Finally, no employees directly through the association. So it will allow us to work on communication, to hold events too. It’s in the plans for the coming years to do OpenRail events.

Walid: That was my next question.

Loïc: That’s it. And also possibly to have a dedicated OpenRail cloud.

There is no shortage of ideas, we will have to prioritise. The goal of the game is not to make money, it is really to coordinate the efforts of the different groups or societies to the common development of digital commons.

Loïc Hamelin

Walid: Can it be a place of exchange? I’m thinking in particular, for example, of open data issues, practices or discussions, contributions, I don’t know, I’m asking myself the question.

Loïc: For the moment, we have ruled out the subject of open data, honestly.

Walid: ok.

Loïc: We have a year of operation. So in one year, we have integrated three new members or four new members. Four new members, that’s it. We have integrated five projects, many more are coming. We’re really focusing on open source for now. Open data will come. It’s like Céline said, we have open source engines. If we don’t put open data gasoline, it will be nothing as a result. You said it better, Céline.

Walid: The last subject I wanted to address, you talked about it a little bit at the beginning, but I’d like us to develop a little bit. You said at the beginning “when I started the project, I was told you’re going to put this project free”. At that point, when you start, do you already have experience? Does SNCF, the group or entities already have projects or experiences, contributions in free software? And if not, is SRD a good showcase? Are there other things going on through this?

Loïc: There’s a project called Tock, which is a conversational robot.

Walid: a chatbot, I’ll put the link to the presentation at Open Source Experience.

Loïc: who is hosted by the TOSIT association. It’s always the same. Simon Clavier is in it. So Tock, it’s about ten years old, I think. It is a production of SNCF at the base. And now it’s being used, adopted and developed by other companies that are contributing. I’m not going to mention them because I might be wrong, but if you tell me that you have a link…

Walid: yes, I have a link to the Open Source Experience presentation, I’ll put it in the transcript.

Loïc: That’s it, so for me, it’s the one that has the most seniority, as a project developed internally. After that, we have been using open source for a very long time, whether for databases, but more for software components, for computer components, but not for business subjects. I am not aware of any other business software that has been developed in-house and that is open source. I could be wrong, but I don’t know anything about it. If not TOCK, if we can consider that it is a business software.

But we feel that I have internal exchanges with some colleagues, we feel that it’s in the spirit of the times all the same.

There are quite a few who are wondering about making their software open source, because they are starting to have industrial interest. Because it will allow you to compare results, it will help you develop faster. We no longer ask ourselves the question of data formats, algorithms, stuff like that. We take them and then we integrate them and it’s over. So at the rail level, it’s taking off quite quickly.

Loïc Hamelin

Walid: There is a difference between creating free software from scratch and open sourcing existing software. It’s still two quite different worlds. Taking an existing software and publishing it is also a very long job that raises a lot of questions about projects, licenses, and the quality of the code. Then you have to animate the community, all that. So what I understand is that OpenRail could also be, somewhere, the receptacle for some of these projects.

Loïc: This is the case for three of the other projects we have integrated. They were closed, they open them to put them in OpenRail, because it makes sense to them. And so, they ask themselves all the questions you just mentioned, because it wasn’t thought of as open free software at the beginning. Unlike what we did, where we started from scratch, we said it’s free, we put developers who knew the open source world very well, so they made the right decisions. And so, for us, it was easier. We have a second project called LibLRS, which is a bit of a spin-off of OSRD, which we detached from OSRD and started as a free software as well, because we think it makes sense to detach it from OSRD, because it can be used in other contexts. It’s a library for converting coordinates between the latitude-longitude and ferro-localization, line-track kilometre marker coordinates, in both directions. So, that’s why we wanted to release it. And we made it free from the start.

We have a project that is being created directly in OpenRail. This is the Germans who wanted to launch a project in OpenRail. And there’s zero lines of code for now. It’s going to deal with software that will help migrate freight wagon couplers, so it’s very business-oriented, but it makes sense because it’s going to affect all of Europe. So why not put it in direct open source? There you go, we have plenty of cases, but you’re right, starting open is much simpler anyway, basically.

Céline: I confirm,

Because we tried to make software open source when I was at Transilien, we didn’t succeed. In fact, since we didn’t manage to make the decision when we started development, we said to ourselves, we’re going to do it, but we don’t really know how to do it, it’s complicated, we have to choose licenses, we have to think about who will animate the community, blah blah, so we’ll do it later. In the end, doing it later is even more complicated. So don’t repeat our mistake.

Céline Durupt

Walid: Are there people who come to see you precisely to discuss your experience within the group?

Loïc: Yes, there are two or three, one outside the group. They came to see me saying that I’m interested in the process, how you did it. That’s what I was telling you earlier. We can feel that things are moving and that we still see the interest of open source for many reasons. It’s not just the belief of public money, public code. It’s not just that. There are other things. There is a really industrial interest in having open code. For efficiency, for obviously code reuse, but not only efficiency in development. Even on a daily basis, having an open code means that you don’t have to ask questions. Basically, we know what’s going on. If it doesn’t work, we can look directly. If you see that you want to develop something that colleagues have already developed, you don’t even ask yourself the question anymore because it’s open. Now, we are able, two polling stations apart, to develop the same thing twice. I don’t talk specifically to the SNCF, it’s in companies in general. Having this approach to opening the code can avoid this kind of problem.

The last word from the guests

Walid: We’re coming to the end. We discussed a lot of subjects. I would like to give each of you the floor for a final word. If you have a message to convey, now is the time. Céline, do you have the honor, do you want to send a message?

Céline: Thank you very much for the invitation. I hope that our conversation has not been too technical, especially on railway vocabulary, which is often the difficulty when we present our subjects. And then I invite all listeners, if you are interested in OpenData and Open Source topics at the SNCF, do not hesitate to contact us. You can come to FOSDEM. If you come to FOSDEM, there is a large part of the team that will be there, so come and talk to us. Otherwise, we will talk through the usual channels. Do not hesitate to contact us. We are always available to discuss all these subjects. There you go.

Walid: Thank you, Yohan, any final words?

Yohan: First of all, thank you for the podcast, for the performance. All that I could say at a pinch,

I would still just like to insist on what open source brings also on a human level, in the sense that integrating a project like this when you start in the activity, is only more to my taste. You meet a lot of things, you meet a lot of people, you learn a lot of things. There are fewer barriers, as a result. It’s easier to appropriate concepts that may seem foggy. I can’t say more than Céline. We are also on matrix.org, we have people who come to see us sometimes. Don’t hesitate, and then contribute, don’t hesitate to come and contribute.

Yohan Durand

Walid: Loïc, your last word?

Loïc: First of all, I too, thank you Walid for the invitation. It’s a bit unexpected to talk about OSRD in a podcast specialized in free software, but it’s a real pleasure. Like my colleagues, do not hesitate to come and see us at FOSDEM, do not hesitate to contact us. I think you’ll put the necessary contacts in the documentation. What more can we say? I’m extremely proud, of course, to contribute to this project, because I think it’s a project that makes sense, it’s a project that has a future. I’ve had it in mind for a long time, but we’ve been developing it for three years. I think we still have great years of development ahead of us and that it will really help promote and improve the rail system. Because that’s really my two goals. And then for those who are interested in software development, I think you just have to look at the code. You’ll see the quality of OSRD’s code, but free software in general. These are people who are extremely demanding. So, you’re going to see some nice code. You should not hesitate to go and look and contribute, of course, because we also need your contributions. There you go, thank you again for the invitation and long live OSRD!

Walid: Thank you to all three of you. I have to admit that the first time I contacted Simon to ask him if I could talk about OpenRail was in 2023. I was crazy, I said to myself it’s so good, we have to talk about it. So I’m happy that we can talk about it. I think it won’t be the last time, in my opinion, that we talk about it because it’s really a subject that fascinates me. Thank you to all three of you for taking the time. We will certainly have other episodes on transport and free software in the coming months. I still have a lot of ideas. Then there’s FOSDEM coming up, so it gave me a lot of ideas. Thank you, see you soon. And for the listeners, as usual, I’m just asking you to share the episode. And then, if you have any ideas or remarks, don’t hesitate to make them. The main channel being Mastodon. There you go, good evening and see you soon for a new episode.

This episode was recorded on January 15, 2025

License

This podcast is released under the CC BY-SA 4.0 license or later.