[LRUG] Field definitions in AR models

[Thomas Buckley-Houston]

Really interesting to hear the viewpoint that AR is just a persistence
layer and not designed to define the domain model. It all makes a lot
more sense now.

@jc I'm interested in your idea about using accessors for each
attribute, I'd like to try that. But yes, it's still not the single
source of truth and so there's the risk of going out of sync with the
DB.

Anyway, lots to think about, much appreciated folks :)