Docker is weird (codeandcake.dev)

So, not only did I not anticipate this kind of success, I did not even have a reference frame for what success might even look like. All I knew was, I had an itch that I had to scratch, and I had nothing better to do :)

None of these things can be transposed today, so I guess I can't answer...

For what it's worth, my current project (dagger) is written in Go.

https://docs.docker.com/reference/dockerfile/

Docker syntax is one of those things that just immediately falls out of my head when I close the editor. I don't care about it at all.

Even better: there are many, many tools that build OCI images without using the Docker syntax. I haven't hand-written a Dockerfile in at least six years.

LLMs have not loved the needle here in my sphere, it was already bouncing off of the zero on the gauge.

edit: this isn't to say the syntax is good or bad (though there are several questionable parts of it) but extremely simple, and concise enough that you can just look up literally any command with great ease. It's surely a large part of its success, for better or for worse.

When I was learning Docker I can't tell you how many people would describe it to me by analogy to a standardized shipping container, and start going off about shipping container dimensions and supersized cargo ships and trucks and railcars... and finish by smiling beatifically, like they had just explained the whole thing in a masterstroke. Let's... uh... not do that. :-D

In a problem set, the student might learn a bit about dockerfiles or docker-compose.yml... but only, I think, so they can try building and running a container of their own!

Yes, I still use Docker. Although my way of using it has changed over time, like many others I assume.

My new project, Dagger, although it doesn't directly compete with Docker, does implement its own container runtime and orchestrator (for CI pipelines), and we recently added compatibility with Dockerfiles and Docker Compose files... So increasingly I find myself using Dagger instead of Docker, in the context of my CI pipelines.

On the other hand, I use docker compose on almost every personal project - like a minecraft server for my kids, home automation, AI hacks, etc. These days I just let my AI agent set it up for me, though :)

I did not start a band, but I did build a nice house, help my family etc.

I still program everyday, working on my second startup, https://dagger.io (please forgive the plug). We have a very active discord if you want to come say hi.

However Moby was a poor choice for a skunkworks project, it involved mostly communications and project management, which requires consistency and scale. Overall it was poorly executed because I failed to build enough support internally ("Solomon's at it again, let him do his thing"). Here I am trying to explain it on Twitter after the keynote, using a picture of a hand-drawn diagram, because nobody in marketing or product was assigned to help. https://x.com/solomonstre/status/855918630915133440

I think the goal of Moby was fundamentally correct: we were trying to replicate the Red Hat model, with well-scoped upstream components (linux -> runc, containerd..) integrated into a commercial product (Docker -> RHEL), with a faster-moving community project in between (Fedora -> Moby).

This was a way for us to clarify that Docker was a product, with a community of users, and Moby was an open source project, with a community of contributors. This clarity was very much needed, because different communities were projecting very different expectations on the brand "Docker", and these expectations conflicted with each other.

In a way, Moby aspired to be what CNCF ended up becoming - the "big tent" where the broader community meets to hammer out the general shape of the container stack. It seems obvious now that a single company can't be the sole authority over such a vast scope. But in the first years of Docker, we really did have that level of influence, because we literally invented the market, and brought the community together in meetups and an ever-growing Dockercon. It's only when Google and Red Hat pushed Kubernetes and CNCF, that our influence over the stack started to wane. Maybe it was for the better, I don't know if we would have been up to the task as designers and engineers. But at least there would have been a consistent design, with a Linus Torvalds figure willing and able to say "no" to stupid and redundant features. For a while I played that role - but then I was replaced by SIGs and committees.

I also feel lucky that people are still using my work, that it endures without me, and that it generates (mostly) positive reactions. There was a time where Docker had a lot of haters, which was hard on me. But with time I've noticed that the negativity gets mostly washed away, and what remains is positive interactions like the ones in this whole thread. So now Docker is this cool experience I share with lots of people around the world, and when we meet we get to talk about this shared interest. I love it.

One design principle that was very important to me, was portability and avoiding fragmentation. I wanted every image name to have the same meaning in every engine, so that when writing a Dockerfile, you could have confidence that it would actually behave the same on every docker installation. In this we disagreed with other platforms (like Red Hat) who sought to "embrace and extend" Docker by changing the meaning of short names. We always fought this to avoid fragmentation of the platform. This did not make sense from Red Hat's perspective, since they were the platform - who cared about fragmenting Docker? That led to some negativity which is probably what you're remembering.

At the same time early Docker users wanted the ability to host their own images without being forced to use our central registry. Using Go's system solved both problems at once.

That day is forever engraved in my memory... Off the top of my head:

I signed up for a lightning talk expecting an audience of 10-20 people, for a confidential demo of our unfinished project that we had so far only showed in 1-1 demos. Imagine my surprise when I learned Pycon does lightning talks on the main stage!

The previous speaker didn't show, so I started early. In the recording you can hear me joke that perhaps I should get his time allocation too - the MC was not amused.

I went over time and had to speedtalk through the end.. The MC once again was not amused and practically shoved me off stage :-D

We had a small sponsor booth at the even, where our small marketing team was prepared to pitch our product (DotCloud) but was thoroughly confused when a horde of people showed up after the talk, to ask about "Docker" - which as far as they knew was the founder's side project.

I drove down to Santa Clara on my newly acquired motorcycle - my very first trip on the freeway. I remember telling myself, "today is the day to take all the risks!"

Python and Go are both memory-managed languages. That is perhaps the hardest part of programming (outside of getting the code to do what is intended).

Go and Rust are both compiled, but Rust requires the developer to figure out all of the memory management before compilation, which slows down development to a crawl.

Almost no company will rationally choose Rust over Go if they know how the development speed tradeoff looks like.

To me the arbitrarily complex docker networking makes perfect sense. Especially in a compose setup. I want my containers to be able to talk to each other, but I don't want them on my local network or even accessible (sometimes) from the local machine. The various internal networks offer isolation of containers from the rest of the network stack. If I have a rails app in a docker compose configuration with PG, Redis, Elasticsearch, etc, in a development environment, most of the time I don't want the service/accessory containers anywhere near anything other than the app's container. And if I want several compose stacks to work together, I just put them on the same named network. I apologize if I misunderstood your point.

The most successful project that we worked on together was "project Piñata" which later became Docker for Mac. It turns out bundling a Linux VM in a native Mac app, and integrating it into the host filesystem, networking stack etc requires lots of magic..

Now, most of the Unikernel team were very active OCaml developers, and created MirageOS among other ocaml projects. While working on Piñata, they simply reached for the tools they were most familiar with.

The first two are just kind of the nature of containers on linux, and applies to any container technology.

Podman does allow you to run it as a non-root user (with some limitations), that was in fact one of the biggest motivations for making podman. This wasn't possible at all in the early days of docker, it wasn't until linux introduced user namespaces, where the root user inside the namespace mapped to a non-root user id outside the namespace that non-root conatiners were possible.

And podman, unfortunately IMO, is also written in go.

:D

Otherwise it’s basically the same and you can continue using your Dockerfiles and Docker Compose files.

Unless you’re doing something really fancy it’s basically a drop-in replacement.

rkt (2013) was CoreOS's alternative to Docker and was daemon-less and failed.

Podman came along relatively late (2018) and benefited from a lot of Linux kernel and systemd improvements since Docker was designed.

So while arguably the client/server model is outdated for container management tools, I'm not sure it's weird. This is my workstation for example:

  schmichael@fedora:~$ ps -e -o comm= | rg 'd$' | wc -l
  67

Now it's possible to create containers without elevated privileges (assuming user namespaces are allowed), but that wasn't possible until long after docker had established the client/server model.