How Docker turned intricate Linux code into developer pixie dust

how spamming hn gives credibility to your company thats the pixie dust :) (it works, too!)

I mount a directory from the host machine as a volume. It works great, though I've only been mounting the volume to one type of container where the UIDs are persistent. Just need to chown the directory once correctly on the host, or include a chown in your dockerfile. It can become problematic if you have multiple containers with variable UIDs, though.

A couple of awkward things with docker:

* Each container is kept indefinitely and it's up to the user to adopt a garbage-collection strategy.

* Images built with a Dockerfile contain the build dependencies even if they're not needed at runtime. That has to be a better way.

* Docker is still single-host. If you want to scale horizontally you need to build another layer on to to run the containers on different machines. But then features like container links don't work anymore.

* Unlike the real physical container it actually matters what's inside the box. You can't run the same container on different CPU architectures.

I believe Docker is great for mass-spawning tons of similar isolated containers, but feels more like an obstacle than help for other isolation/virtualization needs.

I've tried to use Docker to just isolate non-distributed ("singleton") services (in particular, virtualizing a single Puppet Master) and found it just staying in my way and complicating things more than they should be without any gain. But if I'd need to, say, launch a bunch of redundant servers or provide a PaaS-type hosting - I'd probably use Docker.

That is, I guess, due to a hype, Docker just accidentally got marketed outside of its problem domain.

There are a number of ways for handling permanent storage. Volumes on host is a good one. If you create permissions issues with different users that's a problem you created and can easily solve.

Anyone with half a brain and experience with virtualization or dev ops in general can see that the numerous powerful features of Docker provide a superior paradigm.

Your reaction is quite common though. Similar to the reaction people had with Node.js or going back very far to Java. Any new game changing technology that comes along is going to have people dismissing it for various reasons.

You may need to simply say that its not ready because you found some problems that you weren't able to solve such as data storage. This way you can pretend the technology is immature rather than admitting that you weren't able to figure out how to use it.

Other reasons for dismissing paradigm shifting technologies besides just not being able to learn them: status quo bias or sunk costs. Many application developers, once they had dedicated a certain portion of their lives to learning the intricacies of c plus plus were not able to accept that superior application development technologies were invented.