The goal is to build a docker data image that corresponds to
snapshots.  I anticipate that there will be two tricks necessary to
achieve this.

1) Each new release needs to add the incremental sequence files to a
new layer.

2) The sqlite databases need to be a separate layer "above" the
sequence layers.  



Conceptually, this will lead to a flow like the following:

                                                      + t2     
                                      + t1            + t2s    
                      + t0            + t1s    ---->  t1s    
                      + t0s    ---->  t0s      ---->  t0s    
      scratch  ---->  scratch  ---->  scratch  ---->  scratch

Where s indicates a sequence layer, and the undecorated tag includes
the database layer.  Another way to visualize this is:

                           t0              t1              t2
                         ↗              ↗              ↗
      scratch  ---->  t0s      ---->  t1s      ---->  t2s
                                      t0s             t1s
				                      t0s

So, that means each build t<i> is based on t<i>s, which in turn is
based on t<i-1>s.


Naming: Users won't care about the sequences without the
databases. So, I'll create two repos: biocommons/seqs (s tier above)
and biocommons/seqrepo. Each of these will be tagged with identical
tags.


* biocommons/seqs

This will be a data only image.
In order to build t<i>, we'll build from t<i-1>

Since this image is based on scratch, it will have no executables at
all. And, since our update method will require 

The only way to add data will be to attach it to a container that
has an os.

We'll assume that the source data on the local host and mount that
directory into the container.