I happen to do web development and run a Docker course and I will confidently say it's a lot easier based on support questions and overall feedback that I've gotten over the years.
Because the alternative involves a ton of manual steps that every individual needs to take, for example with Python (and this applies to most languages / ecosystems with tiny changes):
- Which version of Python do I install? How do I install it on Windows 10? What about macOS? What's the instructions look like on my distro of Linux?
- How do I handle multiple versions of my programming language or install the language without the root user?
- I need to run Postgres, how do I install that on my OS?
- I need to run Redis, how do I install that on my OS?
- I need to run a specific version of Postgres / Redis for this project but a different version for another project, how to set this up?
- I need to use Webpack, how do I get Node installed and set all of that up
- How can I easily run all of my app's dependent services at once (web + worker + webpack + database + cache) without having to open multiple terminals or discover language specific tools with their own config files
- How can I use environment variables in my app that are loaded from a file?
- My app needs a library that requires a C dependency to be compiled to use it, how do I get that working on my OS?
Now imagine you're just getting into programming and want to build a web app with Flask, Django, Laravel, Rails, Phoenix or any other language / framework. You're going to need to do this for pretty much every tech stack and the learning curve for this stuff is really high when you're a beginner. There's so many unknown unknowns that lead to dead ends causing folks to get really frustrated and just say fuck it and never start.
But with Docker, you install Docker Desktop and technically you can treat it as a black box where everything is on a need to know basis. As long as you can find a reasonable base project to work off of getting started is as easy as installing Docker and running docker-compose up --build, and waiting 5-10 minutes. Now you can focus on learning the important thing (web development using the framework you've picked).
In most cases that works very well. Sure there's some installation related issues along the way but with tens of thousands of folks who have taken one of my courses using Docker, this a huge minority and it's usually something that's easily fixable (I offer support and deal with questions like this once in a while).
All points taken! I just get frustrated because I figure if the developer is taking care of building a Dockerfile or whatever it is with all of that stuff sorted out for me, providing at least those build instructions as part of the README is enough for weird users like me.
Then I can spin up my own VPS/VM with the dependencies myself, which removes the entire black box nature of it.
I get that most people don't want to do that and indeed embrace the black box, but I've never worked anywhere where I could spin up a random application in Docker in any sort of Production or even near-Production capacity and say with a straight face I'd done due diligence in what was going on under the hood. (Maybe people don't use it for this ever though.)
I would probably be more on board if I'd ever had a single Docker experience that was just "install docker and run something" and it worked, but every single time I have run into problems.
Realised I forgot to include my references in my previous post. Again, appreciate these problems are probably old hat for regular users, but for people coming in cold they're just pains in the ass.
Docker on non-Linux OS's is just a VM. You could do the same by starting a linux VM w/ the system environment you're deploying on, and save the Docker overhead.
> Docker on non-Linux OS's is just a VM. You could do the same by starting a linux VM w/ the system environment you're deploying on, and save the Docker overhead.
It's not close to the same user experience.
- If you roll a custom VM you need to first pick a hypervisor. Are you going to use vmware workstation, virtualbox or the native type-1 hypervisor on your OS such as hyper-v on Windows or hyperkit on Linux?
- Are you going to set up that VM using your hypervisor's UI or look into using something like Vagrant?
- What private network subnet are you going to choose when you set up your static IP to your VM?
- How will you get file sharing to work really well (complete with inotify support on Windows)?
- What about starting and stopping the VM for each project? What if you wanted to run a few projects at once with different exposed ports, how will you manage all of these VMs and the resources they've created?
- What if you're running native Linux and still want to isolate everything? Now go back to the first bullet point and add KVM, etc. into the mix
- Which Linux distro are you going to choose to run in the VM? If you're using Vagrant does your Vagrant box's OS support all of the guest features of your hypervisor?
- Now that you have your base Linux environment in a VM, how are you going to set up all of those bullet points I provided in my previous reply. It's back to the drawing board on having to do everything yourself but now on Linux.
A beginner developer is not going to have the same experience doing all of that instead of installing Docker Desktop and running docker-compose up --build. Docker is adding massive value from a convenience point of view and it gets much better once you go beyond your local dev environment too. Now you have a way to easily share your exact project set up process with another developer on any major OS and you can run your containers in the same way on a VPS if you deploy your app there too.
I happen to do web development and run a Docker course and I will confidently say it's a lot easier based on support questions and overall feedback that I've gotten over the years.
Because the alternative involves a ton of manual steps that every individual needs to take, for example with Python (and this applies to most languages / ecosystems with tiny changes):
- Which version of Python do I install? How do I install it on Windows 10? What about macOS? What's the instructions look like on my distro of Linux?
- How do I handle multiple versions of my programming language or install the language without the root user?
- I need to run Postgres, how do I install that on my OS?
- I need to run Redis, how do I install that on my OS?
- I need to run a specific version of Postgres / Redis for this project but a different version for another project, how to set this up?
- I need to use Webpack, how do I get Node installed and set all of that up
- How can I easily run all of my app's dependent services at once (web + worker + webpack + database + cache) without having to open multiple terminals or discover language specific tools with their own config files
- How can I use environment variables in my app that are loaded from a file?
- My app needs a library that requires a C dependency to be compiled to use it, how do I get that working on my OS?
Now imagine you're just getting into programming and want to build a web app with Flask, Django, Laravel, Rails, Phoenix or any other language / framework. You're going to need to do this for pretty much every tech stack and the learning curve for this stuff is really high when you're a beginner. There's so many unknown unknowns that lead to dead ends causing folks to get really frustrated and just say fuck it and never start.
But with Docker, you install Docker Desktop and technically you can treat it as a black box where everything is on a need to know basis. As long as you can find a reasonable base project to work off of getting started is as easy as installing Docker and running docker-compose up --build, and waiting 5-10 minutes. Now you can focus on learning the important thing (web development using the framework you've picked).
In most cases that works very well. Sure there's some installation related issues along the way but with tens of thousands of folks who have taken one of my courses using Docker, this a huge minority and it's usually something that's easily fixable (I offer support and deal with questions like this once in a while).