lol guy makes a fair point. Open source software suffers from this expectation that anyone interested in the project must be technical enough to be able to clone, compile, and fix the inevitable issues just to get something running and usable.
I'd say that a lot of people suffer from this expectation that just because I made a tool for myself and put it up on GitHub in case someone else would also enjoy it that I'm now obligated to provide support for you. Especially when the person in the screenshot is angry over the lack of a Windows binary.
Thank goodness; solving this "problem" for the general internet destroyed it.
Your point seems to be someone else should do that for every stupid asshole on the web?
But will this run inside another docker container?
I normally hate things shipped as containers because I often want to use it inside a docker container and docker-in-docker just seems like a messy waste of resources.
Docker in Docker is not a waste of resources, they just make the same container runtime the container is running on available to it. Really a better solution than a control plane like Kubernetes.
Yeah, it feels like nothing but a little trick. Why would anyone want to actually use this? The exe simply calls docker, it can embed an image into the exe but even then it first calls docker to load the embedded image.
Presumably, they don’t want to write/maintain a shell script wrapper for every time they want to do this, when they could use a tool which does it for them.
> Presumably, they don’t want to write/maintain a shell script wrapper for every time they want to do this, when they could use a tool which does it for them.
How's "packing" cli commands into a shell script any different from "packing" CLI commands into a container?
Calling a container on the CLI is a pain in the ass.
People generally don’t put stuff that works in whatever environment you’re in on the CLI already into contains. Stuff that doesn’t, of course they do.
Having a convenient shell script wrapper to make that not a pain in the ass, while letting all the environment management stuff still work correctly in a container is convenient.
Writing said wrapper each time, however is a pain in the ass.
Generating one, makes it not such a pain in the ass to use.
So then you get convenient CLI usage of something that needs a container to not be a pain in the ass to install/use.
I do that for a lot of stuff. Got a bit annoyed with internal tools that was so difficult to set up (needed this exact version of global python, expected this and that to be in the path, constantly needed to be updated and then stuff broke again). So I built a docker image instead where everything is managed, and when I need to update or change stuff I can do it from a clean slate without affecting anything else on my computer.
To use it, it's basically just scripts loaded into my shell. So if I do "toolname command args" it will spin up the container, mount the current folder and some config folders some tools expect, forward some ports, then pass the command and args to the container which runs them.
99% of the time it works smooth. The annoying part is if some tool depends on some other tool on the host machine. Like for instance it wants to do some git stuff. I will then have to have git installed and my keys copied in as well for instance.
CoreOS had a toolbox container that worked similarly to the one you have (the Podman people took over its maintenance): https://github.com/containers/toolbox
Tip: you could also forward your ssh agent. I remember it was a bit of a pain in the ass on macos and a windows WSL2 setup, but likely worth it for your setup.
Basically the same as Python’s zipapps which have some niche use cases.
Before zipapp came out I built superzippy to do it. Needed to distribute some python tooling to users in a university where everyone was running Linux in lab computers. Worked perfectly for it.
The first of which can be p90 solved by "Okay, type 'apt install dash capital why docker return,' tell me what happens...okay, and 'docker dash vee' says...great! Now..."
Probably takes a couple minutes, maybe less if you've got a good fast distro mirror nearby. More if you're trying to explain it to a biologist - love those folks, they do great work, incredible parties, not always at home in the digital domain.
Is there any alternative way of achieving a similar goal (shipping a container to non technical customers that they can run as if it were an application)?
It feels like there ought to be a way to wrap a UML kernel build with a container image. Never seen it done, but I can't think of an obvious reason why it wouldn't work.
I feel like it's much easier to send a docker run snippet than an executable binary to my Docker-using friends. I usually try to include an example `docker run` and/or Docker Compose snippet in my projects too.
Note to others who might like to write long shebangs: the -S argument there to /usr/bin/env is load-bearing, and if you forget it weird stuff will happen, at least on most Linux systems. I wrote about it a few years ago, based on a true story. https://crystae.net/posts/two-shebang-papercuts/
The beauty of docker is that it is a reflection of how much someone cares about deployments: do you care about being efficient? You can use `scratch` or `X-alpine`. Do you simply not care and just want things to work? Always go for `ubuntu` and you're good to go!
You can have a full and extensive api backend in golang, having a total image size of 5-6MB.
> You can have a full and extensive api backend in golang, having a total image size of 5-6MB.
So people are building docker "binaries", that depend on docker installed on the host, to run a container inside a container on the host– or even better, on a non-linux host, all of that then runs in a VM on the host... just... to run a golang application that is... already compiled to a binary?
I've done both, tiny scratch based images with a single go binary to full fat ubuntu based things.
What is killing me at the moment is deploying Docker based AI applications.
The CUDA base images come in at several GB to start with, then typically a whole host of python dependencies will be added with things like pytorch adding almost a GB of binaries.
Typically the application code is tiny as it's usually just python, but then you have the ML model itself. These can be many GB too, so you need to decide whether to add it to the image or mount it as a volume, regardless it needs to make it's way onto the deployment target.
I'm currently delivering double digit GB docker images to different parts of my organisation which raises eyebrows. I'm not sure a way around it though, it's less a docker problem and more an AI / CUDA issue.
Docker fits current workflows but I can't help feeling having custom VM images for this type of thing would be more efficient.
Yep, then I have some projects that have pytorch dependencies which use it's own bundled CUDA and non-pytorch dependencies that use a CUDA in the usual system wide include path.
So CUDA gets packaged up in the container twice unless I start building everything from source or messing about with RPATHs!
Docker / containers are more than just that though. Using it allows your golang process to be isolated and integrated into the rest of your tooling, deployment pipelines, etc.
It's go; that could be trivially done with a script.
Heck, you can even cross compile go code for any architecture to another one (even for different OSes), and docker would be useless there unless docker has mechanisms to bind qemu-$ARCH with containers and binfmt.
I'd argue that having it in a Docker container is much easier to integrate with the rest of many people's infra. On ECS, K8s, or similar? Docker is such an easy layer to slap on and it'll fit in easily in that situation.
Are you running on bare servers? Sure, a Go binary and a script is fine.
Yep, it's using docker as a means of delivery really. Especially in larger organisations this is just the done thing now.
I understand what the OP is saying but not sure they get this context.
If I were working in that world still I might have that single binary, and a script, but I'm old school and would probably make an RPM package and add a systemd unit file and some log rotate configs too!
I remember thinking that the Visual Basic runtime was unacceptable bloat overhead and now this. Cool work though. Also reminds me of self extracting WinZip files.
I remember those times, as well. There was an amusing (in retrospect) period in late 90s - early 00s where one of the metrics for RAD tools of the day was how large a hello world type app is. Delphi was so popular then in part because it did very well on that metric - the baseline was on the order of 300 Kb, if I remember correctly, and you could have fairly complicated apps under 1 Mb. Visual Basic was decidedly meh on that count because between your EXE and MSVBVM60.DLL, it wouldn't fit on a single 1.44 Mb floppy.
This is useful if you want to share your container (probably something that is prod ready) to someone who knows nothing about docker. An usecase would be, you built a custom software for someone's business/usecase and they are the only one using that particular container.
What do you mean, "requires Windows 11"? What is even "glibc" and why do I need a different version on this Linux machine? How do I tell that the M4 needs an "arm64", why not a leg64 and how is this not amd64?
In other words, it's very simple in theory - but the actual landscape is far, FAR more fragmented than a mere "that's a windows/linux/mac box, here's a windows/linux/mac executable, DONE"
this works for actual compiled code. no vm, no runtime, no interpreter, no container. native compiled machine code. just download and double-click, no matter which OS you use.
On Linux, there would be little to no performance penalty to something like this since Docker is just fancy chroot, re using the same kernel as the host.
But not on other platforms. They are the same but run Linux in a VM.
I'm just as disappointed as I was when I first heard about being able to create 'Self-contained Executable Programs with Deno Compile', perhaps slightly more even as at least that bundled the interpreter.
In all seriousness, Docker as a requirement for end-users to create an executable seems like a 'shift-right' approach to deployment effort, as in, instead of doing the work to make a usable standalone executable, a bunch of requirements for users are just pushed on to them. In some cases your users might be technical, but even then Docker only seems to makes sense when its kept inside an environment where the assumption of a container runtime is there.
I assume extra steps are needed to allow the 'executable' to access filesystem resources, making it sandboxed but not in a way that's helpful for end users?
Why the disappointment with Deno compile? I have not used it but from the website it seems that the end user does not need Deno to be installed. What is the shortcoming you are referring to?
It's not a fair comparison on my part but before reading through the docs some of the initial wording around Deno compile seemed to imply (or I inferred) that a platform native executable would be produced from the process. Wishful thinking on my part I guess.
Other languages like Golang making it relatively easy to build _native_ programs and to cross-compile them makes it a solid choice CLI tools, and I was genuinely hoping that more tooling like that was coming to other ecosystems. Perhaps naive to expect a shift like that for a language that's always been interpreted, but I like when I can run developer tools as native programs instead of ending up with various versions of a runtime installed (npx doesn't _solve_ this problem, merely works around it).
This is super cool — especially for sharing tools with non-technical users or bundling CLIs without asking people to install Docker. Packaging infra-heavy apps into a simple .exe could really smooth out distribution. Curious how it handles startup time and embedded filesystem size.
[1]: https://github.com/NilsIrl/dockerc
I normally hate things shipped as containers because I often want to use it inside a docker container and docker-in-docker just seems like a messy waste of resources.
Can we please go back to the days of sudo dpkg -i foo.deb and then just /usr/bin/foo ?
I haven't tried this stuff, but maybe this is something in that direction.
I don't understand this requirement/specification; presumably this use-case will not be satisfied by a shell script, but I don't see how.
What are you wanting from this use-case that can't be done with a shell script?
How's "packing" cli commands into a shell script any different from "packing" CLI commands into a container?
People generally don’t put stuff that works in whatever environment you’re in on the CLI already into contains. Stuff that doesn’t, of course they do.
Having a convenient shell script wrapper to make that not a pain in the ass, while letting all the environment management stuff still work correctly in a container is convenient.
Writing said wrapper each time, however is a pain in the ass.
Generating one, makes it not such a pain in the ass to use.
So then you get convenient CLI usage of something that needs a container to not be a pain in the ass to install/use.
To use it, it's basically just scripts loaded into my shell. So if I do "toolname command args" it will spin up the container, mount the current folder and some config folders some tools expect, forward some ports, then pass the command and args to the container which runs them.
99% of the time it works smooth. The annoying part is if some tool depends on some other tool on the host machine. Like for instance it wants to do some git stuff. I will then have to have git installed and my keys copied in as well for instance.
Tip: you could also forward your ssh agent. I remember it was a bit of a pain in the ass on macos and a windows WSL2 setup, but likely worth it for your setup.
Before zipapp came out I built superzippy to do it. Needed to distribute some python tooling to users in a university where everyone was running Linux in lab computers. Worked perfectly for it.
Probably takes a couple minutes, maybe less if you've got a good fast distro mirror nearby. More if you're trying to explain it to a biologist - love those folks, they do great work, incredible parties, not always at home in the digital domain.
Pack it in, guys. No magic today.
https://news.ycombinator.com/item?id=38987109
#!/usr/bin/env -S bash -c "docker run -p 8080:8080 -it --rm \$(docker build --progress plain -f \$0 . 2>&1 | tee /dev/stderr | grep -oP 'sha256:[0-9a-f]*')"
You can have a full and extensive api backend in golang, having a total image size of 5-6MB.
So people are building docker "binaries", that depend on docker installed on the host, to run a container inside a container on the host– or even better, on a non-linux host, all of that then runs in a VM on the host... just... to run a golang application that is... already compiled to a binary?
What is killing me at the moment is deploying Docker based AI applications.
The CUDA base images come in at several GB to start with, then typically a whole host of python dependencies will be added with things like pytorch adding almost a GB of binaries.
Typically the application code is tiny as it's usually just python, but then you have the ML model itself. These can be many GB too, so you need to decide whether to add it to the image or mount it as a volume, regardless it needs to make it's way onto the deployment target.
I'm currently delivering double digit GB docker images to different parts of my organisation which raises eyebrows. I'm not sure a way around it though, it's less a docker problem and more an AI / CUDA issue.
Docker fits current workflows but I can't help feeling having custom VM images for this type of thing would be more efficient.
So CUDA gets packaged up in the container twice unless I start building everything from source or messing about with RPATHs!
Heck, you can even cross compile go code for any architecture to another one (even for different OSes), and docker would be useless there unless docker has mechanisms to bind qemu-$ARCH with containers and binfmt.
Are you running on bare servers? Sure, a Go binary and a script is fine.
I understand what the OP is saying but not sure they get this context.
If I were working in that world still I might have that single binary, and a script, but I'm old school and would probably make an RPM package and add a systemd unit file and some log rotate configs too!
Wired: docker2exe.
Inspired: AppImage.
(I'll show myself out.)
I.e. download this linux/mac/windows application to your windows/linux/mac computer.
Double-click to run.
Seems like all bits and pieces are already there, just need to put them together.
What do you mean, "requires Windows 11"? What is even "glibc" and why do I need a different version on this Linux machine? How do I tell that the M4 needs an "arm64", why not a leg64 and how is this not amd64?
In other words, it's very simple in theory - but the actual landscape is far, FAR more fragmented than a mere "that's a windows/linux/mac box, here's a windows/linux/mac executable, DONE"
(And that's for an application without a GUI.)
With dependency management systems, docker, package managers.
MacOS and Windows is closed source and that is of course a problem, I guess the first demo would be universally runnable linux executable on Windows.
The other way around is easier, and already exists thanks to Wine and the ability of Linux kernel to register custom executable formats (https://docs.kernel.org/admin-guide/binfmt-misc.html)
To achieve that you'll need some kind of compatibility layer. Perhaps something like wine? Or WSL? Or a VM?
Then you'll have what we already have with JVM and similar
this works for actual compiled code. no vm, no runtime, no interpreter, no container. native compiled machine code. just download and double-click, no matter which OS you use.
https://github.com/NilsIrl/dockerc
But not on other platforms. They are the same but run Linux in a VM.
In all seriousness, Docker as a requirement for end-users to create an executable seems like a 'shift-right' approach to deployment effort, as in, instead of doing the work to make a usable standalone executable, a bunch of requirements for users are just pushed on to them. In some cases your users might be technical, but even then Docker only seems to makes sense when its kept inside an environment where the assumption of a container runtime is there.
I assume extra steps are needed to allow the 'executable' to access filesystem resources, making it sandboxed but not in a way that's helpful for end users?
Other languages like Golang making it relatively easy to build _native_ programs and to cross-compile them makes it a solid choice CLI tools, and I was genuinely hoping that more tooling like that was coming to other ecosystems. Perhaps naive to expect a shift like that for a language that's always been interpreted, but I like when I can run developer tools as native programs instead of ending up with various versions of a runtime installed (npx doesn't _solve_ this problem, merely works around it).
Except it requires people to install Docker.