Panel: DevRel in and for open source

Amanda: I'm here to introduce our very first session, which is our Google Open Source panel hosted by Erin McKean. I want to tell you a little bit about Erin before we get started. She is a developer relations engineer at Google's, Google's open source programmes. She also runs a nonprofit online English dictionary word nick.com, which I was just telling her is awesome for a grammar nerd like me. Anybody else in the house? Yeah. Thank you. Fantastic. So without further ado, I'm going to turn it over to Erin to talk to us about Google open source. Hey, well, I'm glad

Erin: Y'all are here to learn about open source and DevRel from our expert panellists. So joining us, Victor Gamov, who is heading up developer advocacy at StarTree and he's a Java champion, which I believe you have to do physical fights to be.

Victor: Yes, exactly.

Erin: Yeah,

Victor: Like in the cage.

Erin: Yeah, exactly. The Java cage is terrifying and he's a repeat author of couple books including Kafka in Action. And then we have who, if you've met earlier today, Tamao Nakahara, who is a COO at the moment and a Flux Maintainer. And we also have Joel here who's a DevRel lead, the DevRel lead at MongoDB. Okay. I can see the person in the audience who's wriggling in their chair right now to say MongoDB is not open source. Don't worry, we're going to get to that. But also Joel's coming to us today from Canada and his parents are here. Can we give them a round of applause?

So the next time that your parents compare you to someone that you grew up with, say Joel's parents came all the way from Canada to see him talk and what have you come to for me lately, like that soccer game in middle school doesn't count anymore. So anyway, welcome everybody to this panel. I'm so glad y'all are here. And before we get started, I'd like to kind of gauge our DevRel plus open source kind of, I don't know, levels, right? So if open source is a core part of your DevRel work and strategy, would you raise your hand? Yeah, keep your hand up. If without open source you wouldn't have a job. Okay. How about if open source is a key part of your DevRel work and strategy? It's like how you make things work. Awesome if you don't work with open source at all, but you really, really want to raise your hand.

All right, so it looks like we've got a good number of you here who are working in open source as part of your DevRel work, which is awesome, but if you never think about open source at all, you're in the wrong room. You want to be down the stairs and a little way away. So of course being able to work in and with open source and source available and freely available software is cool, but it also comes with some problems. And I would really like our panellists as the first question to talk about how does open source flavour DevRel make your DevRel role either easier or harder? Start with you, Victor, how about that?

Victor: Given my background in the professional services and consultancy, you always answer. It depends. In some cases it definitely makes my life much easier because I don't need to worry much about if I focusing on the open source things. I don't need to say no, I'll spoil the beans about some of the feature that is proprietary, but also it's actually bad if I will spoil the beans about some enterprise feature that built on top of the open source and things like that. And I was supposed to, not to talk about, but in general, love open source, I was exclusively working with commercial open source companies for the last 10 years. So I would say the upside is higher than downset.

Tamao: I'd say two things. So up until the end of last year, I was at a company called Weaveworks for eight years, which was in the Kubernetes space and had created a project called Flux. So my team was developer experience, so we were everything around that, not necessarily just marketing. So I'd say the two challenges were, so one, we wanted to make sure that our project was really visible and validated. So my team was involved in making sure that the Flux project was within the Cloud Native Computing Foundation, the CNCF, and all the different levels of that. So that was actually a new experience for me. I thought that was really interesting, but you're dealing with an outside org. And then the second part, we still lived within a company with investors and VCs and I was the metrics crank. I had my team doing all these great things and I had to translate that up to making sure that, yeah, all of us C level could prove to the investors that we were actually helping them make money as well.

Joel: For us at MongoDB, open source have been a blessing. We are part of pretty much all the developers ecosystem as being a database. Everybody uses a database and there's so many languages to support. So it's been really good for us that we've had some community people who are working on drivers and making sure that it was easier for people to use MongoDB in their native programming language. So that was a great thing. But on the other hand, we've noticed at some point that there were misconceptions. We were going to events, we were talking to people and they were like, oh yeah, but MongoDB doesn't support transaction. We're like, no, we've been supporting transactions forever. Why do people continuously say that we do not support transactions? And we've noticed that some of those community maintained projects that integrate with MongoDB do not support full set of feature. So it's been both a blessing, but it's also been a curse. Now we've got a lot of misconceptions that are continued because of those open source projects.

Erin: Well, that's a really good segue into the next thing when I think you have to talk about, when you talk about open source, which is the community. Without the community, there is no open source in most cases. And also all technology problems are actually people problems. So maybe Joel, you could talk a little bit more because of course Mongo has a very active community, but also with the relicensing, that was a big deal.

Joel: Of course, in case you didn't know, we changed our licence a couple of years ago. We're now not open source but source available. So it caused a lot of discussions. Of course, there's a lot of pushback in the community. Ultimately, I think most open source projects are really hard. It's really hard to sustain or to make a sustainable open source project and to make them work in the long term. You always need, or maybe not always, there are a few exceptions, but oftentimes you do need that corporate sponsor in the case. In that case, well, MongoDB essentially became the corporate sponsor of MongoDB, I guess. But we still try to include the community in some of the decisions that we were making and really helped the different communities, especially again, when it comes to the drivers, we really rely on open source communities and the people who develop those drivers. So we're really involved with by sponsorship, try to help them if we see that the project, the maintainer is not as active as it used to be because life happens, we'll be out there and we'll offer to help out if they need some help to continue to maintain those projects as well.

Erin: Okay. So I just want to do a little nerd point of order here. Does everybody know the difference between open source and source available? Should I, Hey, there's somebody, so there's an official open source definition and it's held by the open source initiative and they have these criteria that you have to have in order to call yourself an open source licence. And one of them is can you run it anywhere? And there are some licences, business source licences that are like, well, you can run it anywhere except maybe perhaps on a cloud computing platform. Anyway, so you also have freely available, which are things like has anybody ever seen the do whatever the f you want licence? Yeah, that's not actually an official open source licence because it doesn't actually enumerate the rights that you have to use that software. So you don't actually have to be an intellectual property law nerd to work in open source, but it totally helps. So if you would like to be one, there's a tonne of information on the open source initiatives website. So anyway, that was your learning moment for today.

Joel: Thank you for the clarification.

Erin: But most developers often do not care. They want to know, can I use this now? Can I use it for what I want? And only later down the road do they run into trouble if they want to use something in a way that's not possible under the terms of the licence they've picked. So as developer advocates, it can be really important to make your licence terms clear at the beginning to developers because nobody likes an unpleasant surprise after they've already built and launched something. So that's where IP law Nerdery makes a difference in your life.

Tamao: In terms of the community question you mentioned, so I'm part of the flux community and we went through the experience where my company that created it and the term GI ops came out of it, actually died last year. And so of course there was all this nervousness, well, is the project going to live on? And actually two wonderful things came out of it, which was one, I can say now that behind the scenes with our company, major companies like AWS and Microsoft were behind us and financially supporting to make sure the project was successful. But we really struggled to get their people to come to the community meetings. We're like, wait, why aren't you coming? And I guess there was big mentality we're like, whoa, we know that you're going to get it taken care of. So the one good thing was we were in the foundation, all the things that come with the foundation like security audits and all those things validated the project. And then once there was no longer WeWorks behind it, guess who started showing up at the community meetings? Yeah. Now I go every two weeks I go to the online community meeting, the maintainers and developers come and speak and it's just like, yay. I guess this is what it took to get people to step up. So yeah, it just goes to show our technology is solid and the community, lots of other people are now pitching in and giving use cases and speaking. So it's really great. It validates yes, community.

Victor: Two projects that most recently I was working with came from what we call Webscale company called LinkedIn. Maybe use LinkedIn on your day-to-day this day more because of the changes that happened with the introduction of the certain features that were supported by those cool open source platform. And I'm talking about Apache Kafka and Apache Pinot and with Apache Kafka community, it turns out the problems that LinkedIn as a webscale company has many other Webscale companies would help. So they start adopting it there and also most importantly contribute. So at this point, Apache Kafka and Apache Pinot, they have a solid number of contributors from major companies like Uber, Netflix, IBM. And the cool thing about this type of community, there's some really solid innovations coming in. For example, Apache Pinto is a database. Uber contributed a lot of cool stuff for databases like a geospatial index for example, because it's important for them. It turns out it can be important for some other companies who are dealing with the data around geography. So community not only consists of individual contributors, which is very cool, this is how people get encouraged to participate, but it's very cool when companies can invest time to do open source on their dime.

Erin: Thomas, something that you said is a really good lead into something that I think is important to understand is when your product lifecycle or your company lifecycle is anti-cyclical to your open source projects lifecycle. So you were talking about the example of the company was going down, but the project was still kind of going up. Is there more that you wanted to say about how to manage the project lifecycle and the product lifecycle together?

Tamao: Wow, you're getting me into the weeds. So I mean I'm sure plenty of people can analyse how a company's lifecycle succeeds or not. I think that we did have a huge strength that we had this great open source project and we finally, because we were startups, so we had a couple pivots and I felt like we had finally landed on a product line that seemed to really align was moving from open source to paid if people wanted added features, could have been a bunch of things like timing and all that. In fact, the crazy thing was that I was working with the C and CF on a bigger issue that there were still people thinking about all the things that people need to learn to be able to use Kubernetes. And I was like, it's never going to scale. You need to build a thing where it's all abstracted because people in various roles have more important things to do than to take internal Kubernetes workshops or try to train. And anybody who I knew who had started on the journey and got really excited and then suddenly they were like the internal Kubernetes instructor,

Victor: Give ideas to HR people because they might put this as a training. You do compliance, you can do harassment and Kubernetes.

Tamao: But then what happened is they quickly learned how unscalable it was. And so I'm telling all this because our very first product was a SaaS product and we felt pretty sure that we were too early. Kubernetes was really early, and those are all just chess pieces you got to figure out. And so fast forward to basically last year we talked to one of our earliest SaaS customers and the words came out of his mouth. He said, well, what's so great about your product is you abstracted all of Kubernetes away. And I was like, so it's a dance and you have to be able to navigate that. And especially us, we were just really on the front of everything. And so we did a really good run think if you have a company, I think ultimately the company lasted 10 years or more. So hey, I don't think the founders should complain. Feel sad at all.

Erin: Definitely beat the odds if you have a startup for 10 years.

Victor: Can I add something to your question?

Erin: Absolutely.

Victor: So I think there's two things that we need to consider here. The things that Tao brought up is kind of like a SaaS offering. If you're open source, some open source support, some of SaaS offering, it's kind of like makes sense and customers and user cannot expect that. Not immediately the open source version would be available next day even though they want to did this way. But another option is that when you providing some of the open source pieces is a part of your solution, and this is where I would say again in terms of opinions and suggestions, I would say the product managers who will be managing enterprise software, they also great if they would have some sort of experience of managing or at least learning how the open source project were managed. So at the late stage of confluent, we hired a lot of people from the mature, bigger organisations for the product organisation and one of the qualities of those PMs were somehow involved or have experience for dealing with that association because it's not the magic. It's something that you need to learn and experience like how to properly align because people also expecting you're not just simply strip mining in getting the only good cream up on top of your open source and after that repackaged it and resell it, but the expecting you also contribute to something because it's kind of how it works in a good world with good people.

Erin: I do think that's a common anti-pattern in open source where someone comes in who's working on the product side or the commercial thing that's being supported or is part of the open source and they're like, oh great, so you're telling me this community is a bunch of minions who will implement whatever features I want. And they're like, no, no, no, no, no. That's not how any of this works. And trying to say, okay, well everybody has to win for an open source project and the commercial project to be successful, the open source community has to be valued and not just be valued but feel valued for what they bring. And they also need to be listened to and get what they want out of it because a lot of people who are contributing to your open source are never going to be your customers. They don't want to be your customers, but they make value for your customers. But then you have to, and this is what I want to ask you Joel, about especially is so you have to educate the product owners about the value of open source. So how do you do that? What are your tips and tricks?

Joel: It's always hard. I know we had a call recently and we said maybe we could just try to connect Salesforce with GitHub. That would use, that's what you propose, so maybe we can make sure we sell to those people who are contributing,

Victor: Getting hub source and after that, turn them into leads inside Salesforce. Exactly making, here's New York, here's the data for your startup

Joel: Turning PRs into leads. So that might be a solution. No, it is always a challenge to make sure everybody's aligned and trying to sell that benefit. We did a big change in the last couple of months even I'd say where at MongoDB we're all about our own products, making sure that we, because it has the real, it's the source of truth, the only product that we can have out there. And we were trying to crush all of the other mongooses of this world, like all the other ORMs and ODMs, but we completely changed our vision in that regards and now it's all about make sure that we work with the community and bring that value back. And we definitely see a lot of adoption. We've seen an increase in adoption, we've seen there's a lot of metrics that points towards how that effort has been successful.

Erin: Are there particular metrics that you track?

Joel: It's mostly about driver connectivity to the database. So now we can see that people are using whichever driver that they are using or ODM that they're using. So we can see that there's more or adoption based on being with [unclear],

Erin: Is there an internal betting pool? Can you be like, I predict this month this driver's going to have the most connections and can you win money? And if not, can you do that?

Joel: We totally should though, especially that our derail team is organised in language community. So that would be like, oh, you get a bonus if you have a higher, that'd be good.

Erin: I'm just trying to think of all the terrible problems that would come from juicing that veteran

Victor: QC action. Yeah, this is not recorded, right,

Erin: And if you could have biggest increase, you could start your own programming language from zero and then that of course would have a hundred percent increase with just one driver. There's so many ways to gain this. This is awesome. Are there particular metrics you track?

Victor: So not particular metric but particular processes that work great in order to align, say PMs and open source contributors? So in the world of, again, Patrick Kafka, that's community that I'm very familiar with, so I can talk about other things. There's a concept of Kafka improvement proposals where it goes into draught and after that there should be agreement on this. And actually the PMs from the enterprise companies, they can chime in or some people can originate this and after that, if this idea really good and something that they can contribute back will align with the needs of the rest of the community. It's what happens with the tiered storage, for example, implementation because you want to swap different ways how you can store your data and number of these type of keeps or improvement proposal, that's good metric. How these things can work together can align. It's not all about contributors need to come up with all these ideas if you even lead the KIP and maybe someone can pick up and implement this. So that's collaboration.

Erin: So I believe we are contractually obligated to mention AI in every presentation, but I think maybe we can go to the prediction part of the panel where we talk about, okay, well how do we think that AI is going to intersect with open source and source available and freely available materials? And does everybody know that there's actually no agreed upon definition of what open source means for AI yet? You could say open source ai and it's just putting the label natural on a cannabis. It's not

Victor: Funny that does that right? Open

Erin: The open source initiative is now in a discussion stage for what does it mean to be open source and ai. So if you have really strong opinions, go and they're doing this at conferences all over, I think the next session they're holding will probably be around in the open source summit Europe in Vienna in September. But you can contribute to the discussion. But there are lots of things that are part of AI tools like weights, right? Weights are just kind of tables of numbers and there's some disagreement as to what kind of copyright protection they might have. And if something can't be copyrightable, it can't actually have a licence because kind of like saying, oh here I've given you a licence to breathe the air in this room, right? So yeah, there's a lot of confusion going on and lots of things are calling themselves open source when really there's not really an agreed upon definition of what that means yet. So given all that IP not this has been IP nerdery hour, if you were going to predict about how AI is going to change your open source experience, what would you think it would do? Let's start with Tom because you're actually working in a data company right now.

Tamao: Yes. Okay. I have to plug. So the semi stealth startup I echo that I work for, there are two co-founders. One is a big Apache history person in the Hadoop space. We used to work together of course my background and we have another person from old friends come together. So the first thing we're working on, I can plug, is called ai foundry.org. And we're just building the community first to have it be a place to talk about that. And it is so hard to even come up with a name because when you think open ai, you're like, oh, they took that and then they weren't what we wanted them to be, so we could have used the word open and blah, blah, blah. So I'm sure hopefully there's olmo, there's, sorry, Paul Allen AI Institute. There are places that we're starting to partner with and it'll be an interesting couple of years for those of us who care about open source.

And I personally also of course have been starting to plug into various women in AI groups because it's not just a problem about representation that already comes from bad history, but are people actually getting jobs in this space? Are they actually being able to bring both the diversity stuff that they care about, but also those of us who have open source backgrounds? I don't know if we can answer it quite yet, but some of us are at least putting the groups together and moving the conversation forward. Then actually for this conference, when we were going through the call for papers and reviewing them, we were thinking very much, some talks will obviously be like, how do you use AI for your DevRel job? And how do you think about, are we going to be doing DevRel differently when we're working for AI companies? I think again, it's too early to answer, but we'll see where it goes. The ones who know open source, right?

Joel: Yeah, I definitely don't have a prediction. I'm one of those grumpy, old grumpy AI people who don't like ai, but I do definitely hope that open source and AI will bring a lot of transparency in what is going on. There's a lot of things that we're not sure. We dunno where data comes from and it causes a lot of copyright infringement. The other thing I'm hoping is that just like open source really made the entire software industry so much better. Everybody can build on top of each other and construct those huge things that there's no way you can build today from scratch. So I'm hoping that from all of the open source and AI will see a lot more of that, people contributing to each other and building better things. So rather than having everybody trying to build their own things, using a lot of CPUs and kind of destroying the whole planet. So hopefully that'll help and make the whole industry better.

Erin: I just want to do a quick time check. How are we doing? Alright. Okay, now you get time

Victor: And we'll be brief. So in my opinion, lots of things that AI and open source does. I'm trying to look to this one from the different angle from perspective of a generative AI and how knowledge that already inside majority of open source project where they complex like Kubernetes or Linux kernel where people want to contribute or start learning something, but it's so many complexities that it's very difficult to fit in the head of one person, but it would be relatively easy to fit in the head on some AI implementation. So it'll help to better understand the code, maybe automatically research this code from perspective of security vulnerabilities. This is something that, there's certain rules, there's certain things that people trying to prevent and it's kind of relatively easy to teach how to spot this. It's kind of think about this as a, and I said that will be brief, but just go, yes, I think that would be great tool for adopting new contributors. What I'm trying to say in order to understand or what this piece of code it does looks very smart, but what it actually does is not clear because it uses sc. For example, I'm done. So it's my official everything was leading to do a Scala joke on the stage. So thank you.

Erin: I really appreciate the Scala joke. Thank you. Has everybody heard the term imitation intelligence instead of artificial intelligence? That's my new favourite phrase. It's like, oh, okay, well artificial intelligence makes it sound so science and cool. Imitation intelligence is basically like margarine. I can't believe it's not butter. So yeah, I think because I'm kind of oldest dirt, I've seen a lot of these cycles and does everybody remember when everybody was going to have a data lake and everybody was going to have, you couldn't hire a data scientist no matter how much money you threw at them. And I think that we're at the peak hype cycle right now, but there's some interesting things that may become possible, but if you were treating it as a hammer, you're just going to get a lot of bruised thumbs and maybe not as many nails driven in as you would actually like. Do you think we have time for one or two questions? One question from the audience who make it good all? Does anybody have one? Hey Jen, all. So let me,

Joel: We've got all the mics.

Erin: You're doing that so you don't have to answer the question, aren't you Joel?

Audience member: Thank you, so much. Thank you all. Great panel and great stuff. Something I've noticed with open source after COVID had hit the world is I'm seeing fewer younger people entering that space and I'm wondering how can those of us that manage open source tap into that community because they're very different from those of us that are older in terms of how they navigated high school and also events going away for years.

Erin: Okay, can I do a plug for Google's summer of code? We've been going on for 20 some years at this point, and if you have an open source project, what it does is it helps connect you with students and beginners and open source and mentors. So the students get a stipend to work on an open source project and the mentors help onboard the students to the project and it's getting pretty competitive, but it really tries to build those human connections that kind of went away during COVID where you couldn't just show up to a meetup and lurk in the back of the room. And I don't know how we're going to compete with TikTok though. I got nothing on that. I was going to say, you're going to have to start a TikTok all of the projects, but we will see so much more maintainer burnout if now the table stakes for open source is you have

Tamao: To be a TikTok influencer. It's never going to happen. I'll quickly hop on that plug because I had one of my engineers be a mentor in Google Summer code, was working with an amazing engineer and just kept talking every week. I wasn't directly managing, but I just kept hearing about this great engineer. And then at the end of it, of course we wanted to hire her and I was like, Google's going to take her. We're small, but I mean really thanks to how great my engineer was as a mentor, she came to us and I got to work with her. She was so amazing. And so I was just like, I don't think we would've had that if it hadn't been for Google Summer code. And then the second answer, I don't know if it's the answer to your question. I was just having a lunch yesterday with a neuroscience professor here in New York and he's like, I can't get any postdocs because they're all getting tech jobs. So maybe there's a whole different people who would be in academia or like, oh, I just tripled my salary. So who knows things can be moving around.

Victor: One of the things to follow up on top of your question in terms of money, how to get money, it got famous. Don't ask me how in this case, companies actually putting in the companies that I work, they putting this as a job description and many people actually go in there and saying, yes, it's exciting because I get to work on some open source. So imagine you have your house and when no one is there, you have no guests there, there's stuff lying around, there's some clothes, there's some stuff. But when you expect in the guests, your house is always clean and always shine. So that's why people like to have their resume like this and that's why it's very attractive to work on open source project and get paid because the company supports this. So that's another way to do this, maybe incentivize and put this as a kind of like, hey, want to work with something cool and something that other people will be able to see and use. We were supporting some open source projects and we can't pay money for working on open source. I

Erin: Think we just have to make the on-ramps clear to working on open source. Again, we need GeoCities back. Essentially we need to move from a consumption of content model to a creation of content model that's not just short form video. And on that cranky old person note, I just want to thank you all for coming to the panel today. If you have open source questions, hit us all up in the break.

Victor: Great. Thanks.