Design better data modelling / API

[ctheune]

• key/value style or more structured?
• access control to which context can read/write which data?
• types that help coalescing unset values to empty types (e.g. to avoid null handling being special when lists are expected: if 'x' in bar ..: for foo in bar['x']. or for foo in bar.get('x', [])
• secrets
• overlaying
• integrating this with dependency modelling? who provides values, who reads values?

Inspiration can be taken from:

• batou 1/2 with overrides, secrets, provide/require
• puppet/hiera
• chef/pillar
• consul

This would help pave the road to make the modelling more explicit tool-based python and leverage a clean data API that is easy to access in Python and templates.

[ctheune]

This is also interesting in combination with multi-phasic deployments as one phase may need to complete to allow configuring the model for the next phase ... ?