Urbit is this very cool idea where they’re building this cryptographically-secure, peer-to-peer server network where you can just get a server that you can run anything on, but it’s really opaque how to use it; it’s very hard to use. People do use it, and they know this is a problem, they’re working a lot on this… But for us, we wanted to make sure that we solved this problem very well up front.
We saw that people were switching to containers, people were switching to container orchestrators like Kubernetes, and in fact, Kubernetes really seemed like the cream of the crop in terms of container orchestrators… So we made the decision very early on to make our product just completely deploy on Kubernetes. What that means is that when we’re trying to get someone to use our product, if they’re already using Kubernetes, then we can get it up and running for them in like 30 seconds. It’s just a straight Kubernetes manifest that deploys on them. Getting people to that first magic moment with your product, where they’re actually doing something with it – the shorter you can make that feedback loop the better, and the more successful you’re gonna be.
The other very interesting aspect of this is how do you build a company around it, because one of the things that we’ve needed to do, we’ve gotten some developers of the open source community that have come in, but most of the people who do the heavy lifting of Pachyderm development are developers that we employ, and they’re getting paid to do their job, which is a great way to align incentives with money.
For that, you need to have some way that you can eventually make money off of your product. I think for a lot of different kinds of open source projects, I just don’t see any way that this can ever happen. If you’re making an open source BitTorrent client for example, I don’t see how anybody’s ever gonna pay for that. Maybe I’m wrong, I’d love to be proven wrong, but I think a lot of people won’t.
[00:27:51.21] For us, fortunately… Companies, when they invest in data infrastructure, it’s a big investment. If they’re running Pachyderm, they’re probably gonna have 10, 20 engineers who are just using it every single day as a major part of their workflow. In those cases, companies are often very willing to pay for support contracts, because it will just make their developers more effective, it saves money… That’s how we make money right now, we just sell companies support contracts. That means they get to call us on the phone and we’ll fix whatever problems that they have.
The other thing that can work in terms of a business model for open source is if you can turn it into some sort of a hosted model. The analogy isn’t totally perfect, but GitHub is sort of a monetization strategy for Git, right? GitHub itself isn’t open source, but you can see how that works.
We’re planning to eventually build an equivalent for Pachyderm, which is tentatively called PachHub. You can imagine this will be a site much like GitHub, except instead of code repositories there will be data repositories, and there’ll be pipelines that are processing those, and you can see what the entire community is doing with all of this open source data and you can modify what they’re doing with it, and you can contribute back to this open data science community that we wanna build.