Guest: Michel Tricot
Company: Airbyte
Show: Let’s Talk

Airbyte is an open-source data integration engine that helps businesses consolidate data in data warehouses, lakes, and databases. When the company released Airbyte, they did so under the MIT open-source license. That was great for massive adoption, but the company wanted to make sure they were building a sustainable business such that all of the revenue they received could be re-injected into the community and building and making their product better.

According to Michel Tricot, Co-founder and CEO of Airbyte, it’s all about finding the balance with the license. To that, Tricot says, “We want to make sure that nobody can resell Airbyte, but anyone who was building a product on top of Airbyte was leveraging Airbyte internally for data analytics should be able to continue to do it.”

This led the company to consider licenses. On that issue, Tricot says, “We looked at the project and looked at the two pieces that exist.” The first product is the data protocol and the connectors on top of it. Airbyte’s mission is about communicating data integration and all the tooling around that. According to Tricot, “This has to remain MIT. We want that to remain MIT because we are creating a stand for exchanging data. And we want to make sure that everybody can use that freely for forever.”

The other piece, however, is more around like the glue between these connectors and the glue between these connectors is basically what Airbyte is basing their commercial offering on. Airbyte wants to make sure people can continue to use it, integrate it into products. At the same time, that’s the piece Airbyte wants to make sure people cannot just take and resell. One very important thing the company wants to do is to share the revenue that the Airbyte cloud platform generates with the contributors and ensure this is the only way to monetize the connectors.

According to Tricot, Airbyte did look at other licensing models. He says, “a lot of the other licenses are very restrictive in what you can do. The elastic license is the one that is the most permissive, it just focuses on one single use case that is not really possible, which is the reselling of the application. But everything else you can continue to do, you can modify the code, you can distribute the code. You can create a product on top of it, but you cannot just resell Airbyte as open source, like as a project. And that’s the only thing we want to make sure of.”

Tricot also adds that “It’s up to the contributor to decide what license they want to use. Today we encourage people to use MIT.” When those vendors, agencies, or individuals start making revenue from the connectors, Tricot says, “they’ll want to protect themselves and protect their revenue streams from another company that might want to use this connector for free.”

But why not just go full proprietary? Tricot addresses this when he says, “We want to access data in every single silo where it’s available. And if we start having either closed source or just source-available software, that will be a hindrance to the community adopting and wanting to contribute to the project.” Tricot also mentions that it’s hard to find the exact license you need and you really need to be inventive in a world that is focused so heavily on the cloud. For that, Tricot says, “we have a need to innovate about how we think about building a business and, at the same time, giving back to the community and ensuring that they have access to the best of grid.”

The business decision behind this change was centered around supporting anticipated growth. Tricot says, “Today, we’re at 120 connectors. That’s what we’ve built over the past year. Our goal is to be at 200 at the end of the year. And our goal is to be at 1,000 in 2022. And you cannot just fuel this growth, we need to be able to support this growth.”

Tricot is also sure to see a lot more open-source projects making similar moves to converge licenses that work extremely well for the community and makes it possible to build a sustainable business.

One focus Airbyte is concentrating on is what they call the participative model. This is a way for when a connector becomes part of the cloud platform; the company can then share the revenue with the community. This is all about creating a positive incentive that doesn’t just help the company, but the community as a whole to have access to very high-quality SLA connectors. “That is the whole idea behind the licenses. We want to make sure we protect this revenue so that we can feed it back into the community and ensure this high-quality coverage of connectors,” avers Tricot.

The summary of the show is written by Jack Wallen

Here is the rough, unedited transcript of the show…

Swapnil Bhartiya: Hi, this is your host, Swapnil Bhartiya and welcome to TFiR for newsroom. And today we have with us, once again, Michel Tricot Co-founder and CEO of Airbyte. Michel, it’s great to have you on the show again.

Michel Tricot: Thank you, Swapnil it’s great to be there.

Swapnil Bhartiya: If you look at Airbyte, you folks are going through the same kind of soul searching or looking at the business model that many other open source companies are going through in this cloud first, everything as a service world. Because what is happening is that big payers can come in, take your code, it start offering your product as a service and looking at the scale at they operate, there is just no way you can compete with them. It’s not the same old Linux world where Red Hat, SUSE, Canonical and Oracle can fight their… So, it’s not an even playing field. The rules have changed and sometimes I feel and I may be wrong. If you look at licenses, they evolve over time. But I feel that even the most popular open source licenses are, if you look at two bodies FSF or OSI, they kind of feel to come up with the version of license that products, product creators in this cloud world. Of course FSF came up with GPL version three, but the idea was tivoization.

Where you can put everything else. But when I look at a lot of licenses, they seem to product the end user that, Hey, I should be able to do everything that I want to do with this code and not that much about the creator of that product or project. So, that has always been there. If I ask you from Airbyte’s perspective, how do you feel that? We do know that it becomes tricky to run any business, let alone it’s an opensource business, but if I ask you what kind of push did you feel that you’re like, Hey, no, we need to change the license, so I gave you a very broad picture because there are so many moving parts. There is not just one aspect there.

Michel Tricot: When We think about open source and licenses, there are two things that we need to keep in mind. First one is we are building a project for our community. We want to have this adoption and we are creating this standard for data integration. And on the other end, we want to make sure that we are also building a sustainable business so that all the revenue that we get, we can re inject it into our community and into building the product and making the product better. So you have to find the balance with the license.

When, we released Airbyte everything is MIT. That is great, massive adoption, but we also want to make sure that people can continue to use Airbyte the ways they’ve done. And we just want to protect ourselves again, very, very specific use cases. And the only thing we want to do is make sure that nobody can resell Airbyte, but we want to make sure that everybody was building a product on top of Airbyte was leveraging Airbyte internally for data analytics should be able to continue to do it.

And when, we were thinking about licenses, we looked at the project and looked at the two pieces that exist. First one is the standout that we are creating, which is the data protocol and the connectors on top of it. Our mission is really about communicating data integration and all the tooling around that. This has to remain MIT. We want that to remain MIT because we are creating a stand for exchanging data. And we want to make sure that everybody can use that freely for forever. Now the other piece is more around like the glue between these connectors and the glue between these connectors is basically what we are basing our commercial offering on. And we want to make sure that people can continue to use it. You should be able to integrate it into your product and at that point, that’s the piece we want to make sure people cannot just take it and resell it.

And that goes also with our long term strategy about how do we maintain the long tail of connectors, because it’s a very, very hard job and you can build a connector it’s going to work three months down the line it’s going to fail because the external system is broken and you will need to have the community to continue to help on like crowd, on this maintenance. And one thing that we want to do in the future is to actually share the revenue that the Airbyte cloud platform is making with the contributors of these connectors. And that is one reason why we need to make sure that we secure Airbyte cloud and make sure that this is the only place where you can monetize this type of connectors.

Swapnil Bhartiya: Right. There’s so much to unpack there. Number one is that this is not something new we have seen in old days, QT used to have dual licenses, Java used to have dual licenses because they had to maintain and manage that. But they were very clear about it. Here, what I see is that at one point you are saying that you do need MIT license, because you want community support community contribution.

At the same time, you want to product your business as well. Where I see a bit of a conflict there because a lot of companies who have built their businesses, they consume a lot of open source and if that code was released under a license where you cannot repackage it or resell it, those companies would not exist. That totally kind of what the idea of open source, so the idea of open source all the way from MIT to Apache, to FSF, to OSI is that a user can do anything with the code.

Plus, I was talking to Linus Torvalds earlier and I was like, why did you settle down the new GPL version two? He said, even Linux in the early days, they used a lot of licenses, they tried to create their own licenses. But the problem was that you did not even know whether two license are compatible or not. So when you spin your own license and then of course the way we live in this world, there are different licenses. They may not even be compatible. So did you look at other options, also other models as well before using the elastic version two?

Michel Tricot: We’ve looked at other models. And what we thought was a lot of the other license are very restrictive in what you can do. The elastic license is the one that is the most permissive, it just focuses on what one single use case that is not really possible, which is the reselling of the application. But everything else you can continue to do, you can modify the code, you can distribute the code. You can create a product on top of it, but you cannot just resell Airbyte as an open source, like as a project. And that’s the only thing we want to make sure.

I mean, if people want to expose Airbyte for free, they can do it. If they want to use it in their product, they can do it, but not reselling. And that’s why, this license to us was the most permissive. You know, we’ve talked a lot with our community to ensure that, basically we would not be breaking how they were abusing the Airbyte by changing the license. And we tried a few of them. We tried some poly form licenses, which were, which are good, but not for what we wanted to do in our long term strategy for Airbyte. What we want is we are creating a stand out. So we want to make sure that people use us as much as they can with only one restriction.

Swapnil Bhartiya: What is stopping any bit cloud vendor to still take Airbyte and offer it as a free service?

Michel Tricot: Yeah. So there a few reasons the first one is, there is actually one main reason which is connectors. It’s up to the contributor to decide what license they want to use. Today we encourage people to use MIT, but the day we like these vendors or agencies or individual are starting to make revenue out of these connectors. They might want to decide to protect themselves and protect their revenue stream from like also a company that might want to use this connector for free. So at that point, it’s going to be up to our community to decide what they want to do to protect the project and to protect the connectors that they’re building.

Swapnil Bhartiya: Would it be easy to just go proprietary way and say, you know what the source code is available. Because every time I talk to folks and I had actually a show with, Dirk Hohndel from VMware, he is chief opensource officer there. And he was like, just either call it proprietary software or opensource. But when you say opensource, it should be very, very clear that I should be able to do anything with that. So why you did not choose? Hey, there is nothing wrong. Apple is proprietary, actually in most cases, proprietary software is very clear. You do know you cannot do a lot of things, but when you allow them to do things, they’re happy with that. So why to choose the mid part when you do want to restrict a lot of things?

Michel Tricot: Yeah. I think we’re creating a standout. We want to access data in every single silo where it’s available. And if we start having either close source or just source available software for that, that will be a hindrance to the community adopting and wanting to contribute to the project, to build these connectors and have access to the data. And it’s just, there is a long data. Like you have like tens of thousands of different connectors that needs to be built. If we put limitation around who can contribute, how we can contribute, that will hurt Airbyte on the long term and on the creation of the standout. So it’s a balance. It’s a balancing act where we are creating a standout, but we also want to make sure that we are making the revenue, that allow us to power that standout.

Swapnil Bhartiya: If you look at this and you are not alone here, do you see some consolidations will happen eventually where a lot of players who are getting burnt will realize, I mean, there is some efforts going on there and which is not being met very welcoming heart from a larger open source communities. But do you see that at some point, the larger ecosystem will acknowledge this problem and will start to listen to a start of a smaller companies also, and come up with the standardized, another standard, another license that will kind of product them in this cloud, native clouds, first, everything as a service world.

Michel Tricot: Yeah. And I think that’s probably why today, it’s hard to find exactly the license that you need and you need to really be inventive. It’s like we are innovate. I mean, the world is changing around cloud world. How like opensource projects make their way into cloud is, we’re in a new world. So we have a need to innovate about how we think building a business and at the same time, giving back to the community and ensuring that they have access to the best of grid.

So I’m sure that we’ll see a lot more license and a lot more open source project doing similar move in the future it’s just, this project are going to start converging to a license that works extremely well for both the community and for them to build a sustainable company. And like the elastic search, the elastic license is actually a very good step in that direction, in a sense that it keeps a lot of freedom for everybody. And it just limits and protect the project and the company against one very hostile type of behavior.

Swapnil Bhartiya: I will go back to the point of contributors. In most cases, a lot of these contributors, the fact is that unlike some desktop centric or consumer centric projects, where end users like people like me, will go and you don’t send a patch to do that. In most these cases, contributors are either your users or players who are packaging that as a service or a product. When you change a license, some times contributors feel betrayed. They’re suddenly pulled the rug from underneath their feet. You said that the feedback was positive, but you do realize that, there are such cases as well. So what do you have to say to the contributors?

Michel Tricot: So, the thing I want to tell contributor is first of all, everything that like most of our contribution go on the connector side, and this is not something we’ve changed. You can, if you want to use the connector outside of Airbyte you still can. And, but what we’ve done is we’ve really taken a license that is extremely favorable for everybody while making sure that, and that’s why we didn’t go full license on the whole project that I don’t think that would’ve been a positive move. And we just selected with a lot of attention, which pieces we were changing. So they won’t see a change. We made sure that, we are only targeting one thing and the rest people can continue to use, can continue to, they can even contribute to the core platform. The thing that has changed the license, that doesn’t matter, it is still a very, very permissive license. So, where we get most of our contribution, nothing is changed for Contributors.

Swapnil Bhartiya: Right. So anything that is sitting on top of connectors, that’s where you have changed the license. And it might, sometime people get obsessed with that. But even if you look at the Linux kernel, or if you look at Kubernetes, a majority of contributions are coming for the edge of other pieces, not everybody is contributing. So sometime it also that which piece you’re consuming or which piece you are getting involved with. So that also plays with, so thanks for clarifying that because people get offset with the whole picture, but they forget about, they are different components where people are plugging into, or that those are the components they’re consuming. Now, are you worried about any kind of pushback?

Michel Tricot: Not really, maybe one or two people, but I don’t think that we will get any feedback and sorry, any pushback. And we’ve really put a lot of attention. We’ve talked with a lot of members from the committee, like our highest contributors to make sure that and our highest user to make sure that, from their perspective, how would that affect their perception of Airbyte and would that prevent them from using Airbyte and that was a green light all along the way. So maybe, maybe not. I think we’ve made a good choice with that license in the sense that it’s extremely permissive.

Swapnil Bhartiya: What I hear from you. And I feel that you are telling the reality is that these changes are not coming purely from Airbyte but you had a lot of discussion, a lot of engagement with the contributors, core contributors, and you work with them. Because the fact is that the project will not survive without contributors, too. Yes, you need business, but you also need a community around it. So I hear that you work with contributors. So if I ask you after this change, what kind of changes do you expect in the community around hereby? Or what can of changes you would want here?

Michel Tricot: I’m going to explain the community into first one is users. I don’t think anything is going to change for our user community and for our contributor community, there are some logistical thing, like we have a CLA now, but beside that, it’s coding as usual. It’s a contribution as usual. So unless they are, for some reason, they really don’t like this idea of having a non MIT core part of the project. Nothing, nothing really changes. So I don’t expect any changes in behavior and actually the story behind how we want to think about our revenue model in the future, with the contributor, I think it’s going to get a lot of traction around the months as we making that model more real and start engaging more with contributor to share revenues.

Swapnil Bhartiya: What business decision or what drove you to change this license model?

Michel Tricot: Yeah. So today, when you look at what the different companies in industries report in term of number of count were like 150 connector that’s the standard. Today, we’re at 120 connectors. That’s what we’ve built over the past year. Our goal is to be at 200 at the end of the year. And our goal is to be at a 1000 in 2022. And you cannot just fuel this growth, we need to be able to support this growth. And that was one of the reason why we are doing these licenses we want to make sure that we have everything that we need. We can put this connectors in the right phase, in Airbyte, so that we can start doing sharing with committee. We can start incentivize community member to help on the maintenance of this long tail of connectors, because this is extremely, extremely hard. And we’re going to have more than a thousand. We’re going to have a lot more. You have data everywhere. So the scale is enormous

Swapnil Bhartiya: As we were like talking earlier, and you mentioned that this license change kind of came through a lot of discussions with the committee or in a way, not directly from the committee, but through all those discussions. And we also talked about that, how do you want to ensure that the community does not feel betrayed? You mentioned that there are not just one community, there are communities around Airbytes. You also mentioned that you are also working on some model to share revenue. And I also heard a term was participative model. So can you talk about how are you planning to first of all, help community grow so that they maintain that confidence and trust in Airbyte and they grow with your growth.

Michel Tricot: When you’re looking at the long tail of connectors, you have a ton, and the risk that can happen is that someone contributes to a connector, forgets about it, it works for a few months and then it stops working. And so what we want to introduce is what we call it the participative model, which is how can we, when this connector becomes part of the cloud platform, how can we share the revenue in exchange from the community, or from the maintain of that connectors for some SLA like bug fixes, feature ads, and creating this very deep partnership with whether it’s an individual contractor, whether it’s a vendor, whether it’s an agency, like how do we create a positive incentive that helps not just this company, but also the whole community to have access to very high quality, with very good SLA connectors. And that is the whole idea behind, that’s the idea behind the licenses. We want to make sure we protect this revenue so that we can feed it back into the community and ensure this high quality coverage of connectors.

Swapnil Bhartiya: So there you are kind of hitting a lot of birds with one stone. First of all you are trying to kind of eliminate any zombie connectors there is, because that becomes a big problem. And you are also adding incentives for those creators so that they can get incentive. And that also helps Airbyte because you don’t have zombie connectors. So you are kind of hitting a lot of birds with that one stone.

Michel, thank you so much for taking time out today and talk about a bit of a challenging topic today, because talking about licenses in opensource world becomes challenging, but thanks for… First of all, clarifying why you are doing it. And I think most companies today feel this pain, but they really don’t know which direction to go. So maybe these changes will encourage them. And also as we discuss that, eventually we may have a standardized solution to this problem. So the companies won’t have to come up and scratch their own itch every single time. So thanks for sharing those insights. And I will look forward to talk to you again, of course, to understand the model that you’re working on. And of course, to talk more about the cloud that you’re planning. So thanks for your time today. And I look forward to talk to you again soon. Thank you.

Michel Tricot: Thank you very much.