Debian: a Commons of Innovation
I recently returned from Debconf. This year at Debconf, Matthew Garrett gave a talk about the next twenty years in free software. In his talk he raised concerns that Debian might not be relevant in that ecosystem and talked about some of the trends that contribute to his concerns.
I was talking to Marga after the talk and she said that Debian used to be a lot more innovative than it is today.
My initial reaction was doubt; what she said didn't feel right to me. At the time I didn't have a good answer. Since then I've been pondering the issue, and I think I have a partial answer to both Marga and Matthew and so I'll share it here.
In the beginning Debian focused on a lot of technical innovations related to bringing an operating system together. We didn't understand how to approach builds and build dependencies in a uniform manner. Producing packages in a clean environment was hard. We didn't understand what we wanted out of packages in terms of a uniform approach to configuration handling and upgrades. To a large extent we've solved those problems.
However, as the community has grown, our interests are more diverse. Our users and free software (and the operating system we build together) are what bring us together: we still have a central focus. However, no one technical project captures us all. There's still significant technical innovation in the Debian ecosystem. That innovation happens in Debian teams, companies and organizations that interact with the Debian community. We saw several talks about such innovation this year. I found the talk about ostree and flatpak interesting, especially because it focused on people in the broader Debian ecosystem valuing Debian along with some of the same technologies that Matthew is worried will undermine our relevance.
Matthew talked about how Debian ends up being a man-in-the-middle. We're between users and developers. we're between distributions and upstreams. Users are frustrated because we hold back the latest version of software they want from getting to them. Developers are frustrated because we present our users with old versions of their software configured not as they would like, combined with different dependencies than they expect.
All these weaknesses are real.
However, I think Debian-in-the-middle is our greatest strength both on the technical and social front.
I value Debian because I get a relatively uniform interface to the software I use. I can take one approach to reporting bugs whether they are upstream or Debian specific. I expect the software to behave in uniform ways with regard to things covered by policy. I know that I'm not going to have to configure multiple different versions of core dependencies; for the most part system services are shared. When Debian has value it's because our users want those things we provide. Debian has also reviewed every source file in the software we ship to understand the license and license compatibility. As a free software supporter and as someone who consumes software in commercial context, that value alone is enormous.
The world has evolved and we're facing technologies that provide different models. They've been coming for years: Python, Ruby, Java, Perl and others have been putting together their own commons of software. They have all been working to provide virtualization to isolate one program (and its dependencies) from another. Containerization takes that to the next level. Sometimes that's what our users want.
We haven't figured out what the balances are, how we fit into this new world. However, I disagree with the claim that we aren't even discussing the problem. I think we're trying a lot of things off in our own little technical groups. We're just getting to the level of critical mass of understanding where we can take advantage of Debian's modern form of innovation.
Because here's the thing. Debian's innovation now is social, not technical. Just as we're in the middle technically, we're in the middle socially. Upstreams, developers, users, derivatives, and all the other members of our community work together. we're a place where we can share technology, explore solutions, and pull apart common elements. This is the first Debconf where it felt like we'd explored some of these trends enough to start understanding how they might fit together in a whole. Seven years ago, it felt like we were busy being convinced the Java folks were wrong-headed. A couple of years later, it felt like we were starting to understand our users' desires that let to models different than packaging, but we didn't have any thoughts. At least in my part of the hallway it sounded like people were starting to think about how they might fit parts together and what the tradeoffs would be.
Yes, Matthew's talk doubtless sparked some of that. I think he gave us a well deserved and important wake-up call. However, I was excited by the discussion prior to Thursday.
What I'm taking a way is that Debian is valuable when there's a role in the middle. Both socially and technically we should capitalize on situations where something between makes things better and get out of the way where it does not.
I was talking to Marga after the talk and she said that Debian used to be a lot more innovative than it is today.
My initial reaction was doubt; what she said didn't feel right to me. At the time I didn't have a good answer. Since then I've been pondering the issue, and I think I have a partial answer to both Marga and Matthew and so I'll share it here.
In the beginning Debian focused on a lot of technical innovations related to bringing an operating system together. We didn't understand how to approach builds and build dependencies in a uniform manner. Producing packages in a clean environment was hard. We didn't understand what we wanted out of packages in terms of a uniform approach to configuration handling and upgrades. To a large extent we've solved those problems.
However, as the community has grown, our interests are more diverse. Our users and free software (and the operating system we build together) are what bring us together: we still have a central focus. However, no one technical project captures us all. There's still significant technical innovation in the Debian ecosystem. That innovation happens in Debian teams, companies and organizations that interact with the Debian community. We saw several talks about such innovation this year. I found the talk about ostree and flatpak interesting, especially because it focused on people in the broader Debian ecosystem valuing Debian along with some of the same technologies that Matthew is worried will undermine our relevance.
Matthew talked about how Debian ends up being a man-in-the-middle. We're between users and developers. we're between distributions and upstreams. Users are frustrated because we hold back the latest version of software they want from getting to them. Developers are frustrated because we present our users with old versions of their software configured not as they would like, combined with different dependencies than they expect.
All these weaknesses are real.
However, I think Debian-in-the-middle is our greatest strength both on the technical and social front.
I value Debian because I get a relatively uniform interface to the software I use. I can take one approach to reporting bugs whether they are upstream or Debian specific. I expect the software to behave in uniform ways with regard to things covered by policy. I know that I'm not going to have to configure multiple different versions of core dependencies; for the most part system services are shared. When Debian has value it's because our users want those things we provide. Debian has also reviewed every source file in the software we ship to understand the license and license compatibility. As a free software supporter and as someone who consumes software in commercial context, that value alone is enormous.
The world has evolved and we're facing technologies that provide different models. They've been coming for years: Python, Ruby, Java, Perl and others have been putting together their own commons of software. They have all been working to provide virtualization to isolate one program (and its dependencies) from another. Containerization takes that to the next level. Sometimes that's what our users want.
We haven't figured out what the balances are, how we fit into this new world. However, I disagree with the claim that we aren't even discussing the problem. I think we're trying a lot of things off in our own little technical groups. We're just getting to the level of critical mass of understanding where we can take advantage of Debian's modern form of innovation.
Because here's the thing. Debian's innovation now is social, not technical. Just as we're in the middle technically, we're in the middle socially. Upstreams, developers, users, derivatives, and all the other members of our community work together. we're a place where we can share technology, explore solutions, and pull apart common elements. This is the first Debconf where it felt like we'd explored some of these trends enough to start understanding how they might fit together in a whole. Seven years ago, it felt like we were busy being convinced the Java folks were wrong-headed. A couple of years later, it felt like we were starting to understand our users' desires that let to models different than packaging, but we didn't have any thoughts. At least in my part of the hallway it sounded like people were starting to think about how they might fit parts together and what the tradeoffs would be.
Yes, Matthew's talk doubtless sparked some of that. I think he gave us a well deserved and important wake-up call. However, I was excited by the discussion prior to Thursday.
What I'm taking a way is that Debian is valuable when there's a role in the middle. Both socially and technically we should capitalize on situations where something between makes things better and get out of the way where it does not.