This repository contains a series of simple docker image recipes.  These are
meant to be referenced via docker-compose, and built for each project
separately.  No sharing of images.  This provides sanity for each project, as
you don't have to worry about an image changing underneath you.  It also means
that you can be certain you can rebuild on your own whenever the situation
calls for it.

Here are the best practices in these files:

* Generally, any image listed here should run without any external
  dependencies.  No includes from other images, etc.  This makes it easier to
  understand and extend.

* Each Dockerfile is designed to have minimal layers.  This makes rebuilds
  faster, as there are less docker-runs.  It also makes many things easier to
  accomplish, as there is a real shell script that can do sophisticated
  things.

* Images also tend to make use of standard, unmodified debian packages.  Let
  someone else do the hard-work of system integration, while these recipes
  then just do slight tweaks.

* Any files that need to be shared between a host and a container should make
  use of UID/GID being sent in from the host.  The daemon in the container
  should run as the UID/GID, and it's files changed to have that ownership.

  During entrypoint, the target container user is then modified to have the
  correct uid/gid setting.  If there is no target user, then one should be
  added during the image build, generally calling it hostuser/hostgroup.

* Daemons that have complex binary file setups should have those files created
  during image build, then the entire structure placed in a tarball.  The
  entrypoint can then extract this seed tarball, but only if the target
  directory is empty.  This allows for the volume mounting of these
  directories from the host.  Make certain the uid/gid mapping is sane when
  this occurs.