Since the dotCloud platform uses Hipache to route its production traffic, the project is in constant improvement. This document gives some information about what you can expect to see moving in the next future on the project.
For the moment, any value can be passed in the configuration file and if you omit a value parameter, the server will not start.
The idea is to validate the format of the configuration file at startup and to fallback on default value when the parameter does not exist.
While the core of the configuration relies on Redis, it is possible to easily write any configuration adapters to provision the Redis.
A plain text configuration file will be the first one to be implemented: the vhosts and backend will be replaced load from a json file to provision the Redis.
A mechanism will allow to reload the configuration file gracefully while the server is running.
Passive health checks
The current health checking system is active. The backend is checked when a request comes to it. This mechanism has some caveats such as a latency when a backend dies.
The idea would be to implement an passive health checking mechanism to constantly test all backends (using the OPTIONS HTTP method like HA-proxy implements its "httpchk" option).
The current architecture with Redis makes it possible to keep the active health checks system external from the Core of Hipache. This system can be easily run on another machine or a separate process, it will just need to access the Redis to flag a backend as dead.