DinD documentation

[jhamman]

closes #832

[consideRatio]

Wow thanks @jhamman ! That made it a lot clearer to me. I deployed a BinderHub for neurips once and that is all my BinderHub deployment experience, but at that time I didn't understand this but now I grasp this thanks to your documentation!

❤️ 🎉 LGTM

[betatim]

Question for Yuvi and/or Min: are there also security benefits for running builds via docker-in-docker? In my mind one reason to run in DIND mode is that you have better control over what the docker build... command executed by repo2docker has access to in terms of files and resources. But it is a fuzzy concept in my head so I can't write a sentence to include in the docs :(

[consideRatio]

@betatim I find this very challenging, and while working with gitlab's helm chart I've considered related security concerns about Docker overall to build and run things for users in CI pipelines.

Here are some of these concerns written down by GitLab's team:
https://docs.gitlab.com/runner/executors/kubernetes.html#using-docker-in-your-builds

[jhamman]

Bumping this so we don't loose too much momentum. @betatim / @consideRatio - should we add some comments about DIND and security? I would need input from you on what to write.

[consideRatio]

@jhamman I'm not knowledgeable enough about these aspects yet to suggest something =/

[betatim]

this and the next two lines need to be indented. Though they are the same as the default so we probably don't need them at all?

[jhamman]

I've fixed the indentation. I think it is helpful to show the default the image specification here though as this is the sort of thing someone may want to customize.

[betatim]

Suggested change

	for you to tweak. Below is a few pointers for how to configure some of the most
	for you to tweak. Below a few pointers for how to configure some of the most

[betatim]

When PRs go stale feel free to bump them by following up on the questions/discussion points by suggesting a resolution that seems to follow the consensus. I think a frequent reason for why things go stale is that there is no immediate actionable thing and no one is trying to get to one. I think it is best if the PR proposer tries to do the driving by suggesting a next actionable item.

For now I would skip the security question and maybe @jhamman can create a new PR addressing/adding thoughts on that. That way we can get this PR merged.

[jhamman]

@betatim - thanks for the review. I've updated things a bit and this should be good to go in.

should we add some comments about DIND and security

I don't know. I'm not familiar with this issue so I'll leave this for the next contributor.