Docker
computer.rip
computer.rip
They've raised a tonne of capital, and it probably looked sane at the time. And now they're grasping at straws trying to figure out how else they can turn this into revenue.
A lot of the recent narrative has been worded like pivoting into a glorified webhost was their evil plan all along. It's not, it's an act of desperation.
Every year at DockerCon, there would be flashy announcements that went nowhere. As a developer, those years from 2013 to 2017 were both super exciting and super frustrating. Everything started falling apart when Docker (the project) got split into Moby for open source and the rest went commercial. Docker started to sell Docker Swarm (the original), only to kill it a year later with a new Docker Swarm (what we have today). Then, Kubernetes started growing traction, leapfrogging both Docker Swarms, Mesos, and others in adoption. They never had a cohesive commercial plan. Just lots of empty promises and burned bridges.
When I think of Docker (the company), I feel bitter about all the projects they killed in their attempt to own the market. I love using Docker (the software), but the company's just one big disappointment.
I don't think you have Docker without VC money. I think at best you have some LXC-lite project that is intertwined with GCP/AWS/Azure but unfortunately, I don't think you get something as quite polished as Docker in the same timeframe without VCs pouring millions to hire people to work on it.
Where I am sympathetic to Docker is Docker wouldn't have worked as your standard Open Source project; it took a ton of paid engineering hours (and you can argue X% was wasted on projects that went nowhere, but that's a given in any org), to get the software and infrastructure right, and if they had tried to charge developers they would have gotten nowhere. Even now, where people are keenly aware of the value of Docker, trying to monetize it is met with tears and angry blog posts.
I think Swarm was the plan (as the money has always been in providing infrastructure), Google just had more developer sway (which also killed Mesos).
The way I see it are there probably a ton of services you could build with the right team and right amount of people that could make an even greater impact than Docker on productivity and they would never be built because it would be way too difficult to monetize and the people with the talent to build it are going to get paid more working at FAANG to do any sort of OSS approach.
There's a big difference between creating value and capturing value. Probably thr biggest single issue with capitalism in general.
They were always in a race to reinvent everything the Docker way. Which is, sort of, what you need to do if you want to become an enterprise software company, if that's where you think the money is. It's pretty much exactly in the footsteps of VMware. But to make that work, they would have had to work much closer to Windows, which probably was even harder than the Linux community.
Docker couldn't have existed without VC money. Free hosting was what made them hugely popular. Docker was a VC productization of containers.
But they weren’t paying back fast enough, and K8s was making waves. Better make money fast while you can. Time for the squeeze play so the VCs can win.
If they had been allowed to grow at a more natural rate maybe it would all be fine. If they were allowed to be happy with a 40% share of a big future market, things could be great.
But that’s not the VC-shoot-for-the-moon way.
"Docker is also a company, and companies are expected to produce revenue." That reads like wages, office rent and perhaps a boat for the founders. But they can't be that humble anymore, not after burning through a quarter billion in VC money. (even if those numbers seems surprisingly harmless compared to some more consumer oriented VC bonfires)
Now it's their job to make those investments winning bets or ruin the company trying. All that free image CDN convenience much of the docker revolution was built on? Too good to be true, effectively a loan from VC betting on getting people hooked.
> be the cashcow that turned
> that into cashflow. And then
> k8s happened.
We are still using Docker Swarm in production. It seems to be working fine so I always wondered why it never took off. But I am not a Devop. Can sb please give some insight on why kubernetes took off instead and why Dock Inc. failed with its cloud product?
There are quite a few optional tools built on the k8s model, like service meshes or tools to orchestrate a postgres cluster.
Swarm is a small piece of that, the scheduling system, and you have to run it yourself. It's really not comparable.
IMHO, Google did something smart/evil here. Google did not use the docker-compose style yaml files for k8s configurations. This turned into a big deal. If you have a docker-compose file and want to run it in k8s - you cannot. So you need to make a decision: go with docker-compose yamls which has no place to be hosted except swarm OR go with k8s yaml style and be able to run it on aws,gcp, etc.
I rather have to deal with WebSphere 6.1 yet again, than a poorer replication of its developer experience with lesser tooling.
I could see that I was being too abstract, so I distilled it -- and I more or less begged him to not take any more money. (Sadly, Silicon Valley's superlative, must-watch "Sand Hill Shuffle" episode would not air until 2015, or I would have pointed him to it.) When they took the $40M round -- which was an absolutely outrageous amount of capital to take into a company that didn't even have a conception of what they would possibly sell -- the future was clear to me.
I wasn't at all surprised when the SVP washouts from the likes of VMware and IBM landed at Docker -- though still somehow disappointed when they behaved predictably, accelerating the demise of the company. May Docker, Inc. at least become a business school case study to warn future generations of an avoidable fate!
- Bryan Cantrill, https://news.ycombinator.com/item?id=28460504
A 13px font for paragraph text is nearly hostile. It's not that legible to people with perfect eyesight, but then it's not at all legible to anyone with imperfect eyesight. It's like saying you don't care if anyone who reads your blog would struggle doing that. And given how very simple it is to change, it's kind of insulting, specifically given how many years usability has been a thing.
10 years ago I wouldn't have written this comment. But now this isn't how you behave if you have an audience.
And yes, I do believe that Hacker News needs a bigger default font.
(I'm getting old. You will too!)
Besides:
- Nearly if not all desktops allow to scale. If your eyeseight is that bad you should set it at desktop level anyway.
- All browsers and terminal emulators allows one to use his own defined fonts and size.
- Nearly if not all browsers and terminal emulators now allows you to zoom dynamically for that odd website and keep that preference.
- Firefox has reader mode, I guess similar extensions exists for most browsers.
> And given how very simple it is to change, it's kind of insulting, s
Changing by which size? 16px, 32px, 64px? There is no single form of universality regarding eyesight. And I would argue that if your eyeseight is bad, the solution are prescription glasses, not websites with huge fonts.
They aren't as the rendering of a website is ultimately always controlled by the client/visitor.
Any website whose semantic is only made of text is infinitely more accessible than another that would use bigger fonts but would include images or content that only render if javascript is activated.
And ultimately, they have every right to make it the way they want.
> Washington Post are using 20px (and those are usually longer reads)
In the full text maybe, I find much smaller fonts than those on computer.rip in metadatas and subtitles.
And yes, the 20px font applies to the full text. I'm not arguing that smaller fonts should be banned, things like metadata don't need to be as large. But the main content of the page should be larger and comfortably readable.
I've been wearing glasses since I was 6 or 7y old.
This true in a technical sense, but not in a practical sense. If you want me to stay on your website, you'd better put in the effort to make it usable and readable. It's not the user's job to design the website, it's the job of the damn website designer!
Firefox even lets you set a minimum font size. And there is an option to stop websites overriding your choices which helps with sites like HN (but not the one you are complaining about) which explicitly set a small font size.
On the plus side, it would perhaps give enterprises more confidence about their build pipelines remaining dependent on Docker Hub, maybe even being more comfortable paying for it.
On the flip side, far too much of the dev ecosystem would depend on Microsoft, the famed supervillain of open communities. EDIT: With that sense in mind, I am indeed rooting for Docker Inc. to succeed.
You can always use Podman. We already have fully OSS solutions in the container space.
Often, there isn't time for an image to be pulled from a mirroring serving.
There again there's the question of finding image sources you can trust to be updated and secure. Docker Hub has a "Docker Official Image" tag for critical base images that are managed by each community.
Re. finding the sources, a central namespace may still be useful, hence the use of caching proxies.
My suspicion is that since you pay for client connections, they don't want you running a single bastion client and having your real clients connect via that. But it's annoying, and if you really wanted to do that, you only have to edit the script, or set it back on a schedule/after starting up the client.
It does… in the same sense that docker (the program/tool) does, that is: both are not container runtimes (such as containerd which docker uses, or runc and crun which are the options typically used with podman) but a container management tool that control a container runtime. So you would indeed use podman to create a container just like you would with docker.
As for building images, buildah is the tool most used in the podman community for that. and yes, both podman and docker can handle containerfiles (what is/was called "dockerfiles" in the docker world)
> need something for my weird windows+wsl setup
oh, well, uhm, my condolences for that. Luckily I never had to use that for containers, but a quick look on the podman homepage tells me that they also offer a virtualized WSLv2-based distribution for Windows users: https://podman.io/getting-started/installation.html#windows …And of course, there is Podman Desktop if you want something more click-UI-based than the command line podman (never really tried it though, so I can't really say if it's good or not): https://podman-desktop.io/
I do not miss Docker at all considering that I can copy paste almost every docker invocation I see online and have it run flawlessly with Podman. Unfortunately my workplace will probably never even consider trying out Podman as a replacement to Docker. I wonder if someone here has a nice anecdote of using Podman successfully at their workplace.
https://github.com/containers/podman/issues/17580
Buildkit needs a full API implementation
Does that hint that the model of sending around megabytes-to-multigigabytes of VMs is inherently too expensive to maintain as a backbone for an awesome tool?
For the same reason, I wonder why provides Maven Central and NPM repositories, whether they will do it for free, but at least those are billions of small jars, not hundreds of thousands of gigabyte-sized VMs.
Yes. I think automated build pipelines running 24x7 that could request even for the oldest version of a sizable image without caching at their end is part of the issue. There was no limit that I'm aware of on the no. of tags/versions per image or per OSS account on DockerHub, so just like package repositories, effectively every image had to be available forever and each image was of significant size. I don't believe storage and network tx costs have reduced at the same rate as increase in adoption of build pipelines and automation.
Same issue exists for apt, NPM, Maven, PyPi, or any other repository, but yes, the storage requirement should be significantly smaller.
Aside: Because Java has been around for so long in enterprises, many have learned over time to set up registries internally - a combination of wanting to host private packages securely on prem and protecting from downtime, supply-chain attacks. JFrog Artifactory is pretty commonly seen. However, IIRC npm registry was not easy to self-host on prem in the early days, and many enterprises had their private packages hosted on npm.
FOSS: "Never thought I'd die fighting side by side with Microsoft."
Microsoft: "What about side by side with a friend?"
FOSS:
Who would imagine that 20 years later, Microsoft would become yet again a Java vendor, with their own OpenJDK based distribution and upstream contributions to ARM JIT, and escape analysis improvements?
My guess: Because not all good ideas are profitable. Especially in software.
I read most but not all of the article, so if I missed this already being stated, that’s egg on my face.
What is the alternative that is better? The ability to have layers that build on top of each other and can be cached is a big feature... what alternatives provide that and are better?
Taken to the extreme, look at something like Nix (or conda, come to think of it). Why can’t I just have one copy of a package of a given version shared by all containers, if they all want that package? Unix file systems should be great at that kind of composibility; that’s the advantage of a unified tree instead of a tree-per-source. But in the docker model, you’re stuck with a stack.
My ideal image definition is a hybrid between docker’s immutable hash-addressed image layers and an fstab file to describe how and where to mount them all.
Unfortunately, those behaviours are mutually exclusive with transparent layering.
It's certainly possible to build a file-system whose behaviours are compatible with that kind of transparent layering - Plan9 was built on exactly that model, for example - but then it wouldn't be a POSIX-compliant filesystem anymore.
The promise of Docker was that you'd be able to deploy your existing applications in a more reliable, repeatable way, but that breaks down when you have to tinker with your application's file-handling code, or jump through extra hoops to flatten the layers of your container's filesystem image.
* The general idea of mixing together filesystems+folders to achieve re-use/sharing/caching.
* The "Dockerfile" approach to this - with its linear sequence of build-steps that map to a linear set of overlays (where each overlay depends on its predecessor).
The "Dockerfile" approach is pretty brilliant in a few ways. It's very learnable. You don't need to understand much in order to get some value. It's compatible many different distribution systems (apt-get, yum, npm, et al).
But although it's _compatible_ with many, I wouldn't say it's _particularly good_ for any one. Think of each distribution-system -- they all have a native cache mechanism and distribution infrastructure. For all of them, Dockerization makes the cache-efficacy worse. For decent caching, you have to apply some adhoc adaptations/compromises. (Your image-distribution infra also winds up as a duplicate of the underlying pkg-distribution infra.)
Here's an alternative that should do a better job of re-use/sharing/caching. It integrates the image-builder with the package-manager:
https://grahamc.com/blog/nix-and-layered-docker-images/
Of course, it trades-away the genericness of a "Dockerfile', and it no doubt required a lot of work to write. But if you compare it to the default behavior or to adhoc adaptations, this one should provide better cache-efficacy.
(All this is from POV of someone doing continuous-integration. If you're a downstream user who fetches 1-4 published image every year, then you're just downloading a big blob -- and the caching-layering stuff is kind of irrelevant.)
I disagree, I believe Docker /was/ revolutionary. And I feel like I see heavy technologists make this sort of dismissal based on technical points too soon. From a technical perspective, it was arguably evolutionary — a lot of people were poking at LXC and containerization a long time before Docker came around — but from a product perspective it was surely revolutionary.
I used to joke, in my own experience building a business in the DevOps space, that you’d spend 2 years building a globally distributed highly scalable complex piece of software, and no one would pay for it. Then you slap a GUI on it, and suddenly someone is willing to pay a million dollars for it. Now, that’s mostly tongue in cheek, but there is a kernel of truth to it.
The kernel of truth is that the technology itself isn’t valuable; it’s the /humanization/ of a technology, how it interfaces with the people who use it every day.
So what Docker did that was revolutionary was take a bunch of disparate pieces, glue them together, and put an incredible user experience on top of it so that that technology was now instantly available in minutes to just about anyone who cared.
At some point in the article, the author says it’s maybe something about a “workflow.” I’m… highly biased to say yes, absolutely. One of my core philosophies (that became the 1st point of the Tao of the company I helped start) is “workflows, not technologies.” When I talk about it, I mean it in a slightly different way, but it’s highly related: the workflow is super valuable for adoption, the technology is to a certain extent, but less so.
Technology enthusiasts (hey, I’m one of you!) usually hate to hear this. We all want to think you build the best thing or a revolutionary thing and then it just wins. That’s sometimes, but rarely, the case. You need that aspect, and you ALSO need timing to be right, the interface to be right, the explanation to be right, etc. Docker got this all right.
(Now, turning the above success into a business is a whole different can of worms, and like I said in the first paragraph, I don’t plan on commenting.)
For the author: I don’t mean any offense by this. I mostly agree with the other points of your post. The “FROM” being revolutionary I was nodding quite vigorously. Being able to “docker run ubuntu” was super magical, etc. I mostly wanted to point this because I see MANY technologists dismiss the excitement of technologies purely on the basis of technology over, and over, and over again, and the sad thing is its just one part of a much bigger package.
I'm not sure that we really disagree, but I wrote this sort of late and I also think I wasn't entirely clear. The point I was trying to make is that the "container runtime" part of Docker is a lot less important than the tooling they put around it, and they made Docker Hub a very core part of that broader ecosystem.
Apple in a nutshell.
If you want a cash cow, you don't want a technology-focused project or a mere company, but something between a luxury lifestyle brand and a cult.
We used to buy the whole vertical experience, from hardware, OS integration and peripherals.
The clone market, which IBM failed to prevent, kind of broke this down.
Yet it is no surprise that OEMs nowadays try to bring this model back, not only because of Apple, also because it is the only means to differentiate themselves, while recovering the margins that have been lost all these years.
Instead of "user", the more fitting word would have been "customer", though even that would only have been a rough one-word approximation for the concept that could in the Apple brand marketing context better be described as "person (to be) suggestively conditioned to be a follower/fan who is content with being kept in a walled garden and milked as much as possible in order to be part of the circle and to continually FEEL the EXPERIENCE".
So yes, "EXPERIENCE" is key, but "user experience" in the sense that "user" has in development/tech or even specifically in UX is only a smaller part of that.
I get that NPM packages are smaller than docker images typically.
You can run an in-house npm repository with that sold by Npm Inc.
I don’t know how sustainable it is but that’s probably one of their cash cows.
Either that or you vendor in your dependencies.
[1] https://smalldata.tech/blog/2023/03/17/setup-a-private-npm-r...
At current average SaaS revenue multiples (6.7), Docker is on the cusp of Unicorn status.
It's weird to read comments about "poor, sad, dying Docker" given how ridiculously successful Docker's Desktop licensing scheme is.
https://devclass.com/2023/03/24/docker-subscription-revenue-...
Solaris containers didn't have anything like Docker Hub nor was setting them up as easy as "docker run".
(Posting as ex-Solaris guy)
My introduction to containers was in HP-UX in 1999, via the HP Vault infrastructure.
Tru64, while short lived in the market, also had similar ideas.
And then there is the whole linage of mainframe and micro computers from IBM.
About the only real negative of Nomad is that it doesn’t have the mindshare that k8s does, so you don’t see the amount of developer engagement in extending it the way you do in the k8s SIGs. Also, being an expert in Nomad doesn’t give you the same number of career opportunities, and on the other side - there aren’t umpteen thousand nomad SREs the way there are with k8s - so getting someone up to speed can take a couple months (but this system is very well defined, well documented, and small enough that any half talented engineer can master it very quickly)
Nomad does have the very important advantage that Hashicorp stands behind the product - so if anything goes awry, you’ve got a support team and escalation that will jump on and root cause/resolve any issue, usually within a matter of hours and even in the really squirrelly cases (that you are only likely to see when when you are managing many, many thousands of nodes in a cluster) within days.
Since the last couple of versions there has been native Service Discovery in Nomad which works pretty well.
ECS, nomad or even autoscaling VMs are much easier to deal with when they fail.
You change your docker-compose, push and we detect via webhook and deploy. Logs, metrices everything from command line with bubble tee or something.
Most companies have brilliant engineers and shortsighted, incompetent out of touch product teams.
So making a bad decision is bad, but admitting it was a bad decision and reversing it is also bad?
If they would mandate a registry then many more people would host their own and take the load off of their system.
But no, they want to have it all. And yea, they can. That I don't care about.
But you can't make a change, and walk back from it, and expect people to be happy, given the story that came before all of this.
Currently the situation is that a lot of people will think twice before generating any dependencies with free teams, or perhaps Docker altogether.
Probably, especially from those who think in the long term. Like those who builds things for themselves. They don't like to read news from Docker every day and keep in mind that their project images and even base images can just disappear overnight. It's too expensive for them to track docker decisions. It just takes resoures.
Making bad decisions is not bad at all. Losing trust is.
There are also good and bad ways to apologize and change your mind. I don't have an opinion as to whether or not Docker's apology and reversal were done well, but I think it's fair that some could believe they weren't.
Not only that, but it was actively encouraged by all Docker fanbois to pull as soon as you can. When I saw Watchtower the first time I was just speechless.
Though IMO they had a chance at getting money long before that debacle: https://news.ycombinator.com/item?id=34377674
But that achieved their goal too? They wanted to reduce loses from bandwidth costs, that works by either making the users pay, or use less bandwidth.
Yes, this is the default assumption.
Wild to me that Docker inc. Doesn’t just charge to pull prebuilt images. Just send docker files!
They had guaranteed cost (hosting & serving a bunch of heavy data) and no obvious monetization play available.
How can the decision and reversal both be boneheaded?
> Docker Hub, though, may yet be Docker's undoing. I can only assume that Docker did not realize the situation they were getting into. Docker images are relatively large, and Docker Hub became so central to the use of Docker that it became common for DevOps toolchains to pull images to production nodes straight from Docker Hub. Bandwidth is relatively expensive even before cloud provider margins; the cost of operating Docker Hub must have become huge. Docker Inc.'s scaffolding for the Docker community suddenly became core infrastructure for endless cloud environments, and effectively a subsidy to Docker's many users.
I'm not sure why they couldn't have been a bit more aggressive about monetization from the start?
DockerHub could have been free for an X amount of storage, with image retention of Y days by default, with Z amount of traffic allowed per month. The Internet Archive tells me that they got this half right, with "unlimited public repos" being where things went wrong: http://web.archive.org/web/20200413232159/https:/hub.docker....
> The basics of Docker for every developer, including unlimited public repos and one private repo.
For all I care, Docker Desktop might have just offered a CLI solution with the tech to run it (Hyper-V or WSL2 back ends) for free, but charge extra for the GUI and additional features, like running Kubernetes workloads. BuildKit could have been touted as an enterprise offering with immense power for improving build times, at a monetary cost.
Perhaps it all was in the name of increasing adoption initially? In a sense, I guess they succeeded, due to how common containers are. It is easy to wonder about these things after the fact, but generally people get rather upset when you give them things for free and later try to take them away, or introduce annoyances. Even if you cave to the feedback and roll back any such initiatives, the damage is already done, at least to some degree.
I still remember a piece of software called Lens one day starting to mandate that users sign in with accounts, which wasn't previously necessary. The community reacted predictably: https://github.com/lensapp/lens/issues/5444 (they also introduced a subscription plan later: https://www.reddit.com/r/kubernetes/comments/wakkaj/lens_6_i...)
That said, I self host my own images in a Nexus instance and will probably keep using Docker as the tooling/environment, because for my personal stuff I don't have a reason to actually switch to anything else at the moment and Docker itself is good enough. Podman Desktop and Rancher Desktop both seem viable alternatives for GUI software, whereas for the actual runtimes and cloud image registries, there are other options, though remember that you get what you pay for.
If it was paid, even a small amount, that’s a hurdle for people. Plus people avoiding it would have created more/stronger competing products as they had more incentive.
Get a ton of users then try to monetize later is a very common SV play for VC backed companies.
I'm not sure there is that much money in running a glorified specialized S3. Charging for disk space is terrible - the people who have money would probably just setup a private repo on S3 where it's cheaper and the people who don't aren't going to pay you.
For the amount of money they raised I don't think that would have been a convincing story to tell.
Docker as a company may be a joke, but I don't think the software will be nearly as nice to use without them. I think it's ridiculous that so many asshats are jumping on the hate Docker (the company) bandwagon without understanding how much they have been taken advantage of by the big players who can absolutely support them, but choose not to.
Sometimes I am so disappointed at how much ego still exists in tech. We're supposed to be more educated than the folks who came before us, yet we're doing a worse job.
As much as I do not condone said big player's actions here, the whole system just doesn't reward "doing the right thing" as a general rule. If the licensing allowed it, they were in their right to do it. Even if they did do their right thing and support them, their competitor may not have. The morality element just doesn't have much weight the way things are.
I don't know if it needs to be held back, but the product needs to have innovation or it's going to feel like features are being held back and put a sour taste in the community's mouth.
I get the outrage. If it's pointed at Docker's inability to productize, then bravo, but I think AWS and the other big guys deserve most of the blame for taking tons of money on the table without contributing much to the state of the ecosystem. It's just tragedy of the commons with more steps.