scorsh/notes.txt

## structure

- we read the list of workers from the config file. Each worker
  record consists of

  - a list of repos/branches
  - a folder where the configs and logs are kept
  - a logfile
  - a tagfile with the definition of tags
  - a list of keyring files

## master logic

- in main() (master) we create a worker for each worker record,
  maintaining a map of which worker can process commands for which
  repo/branch

- The spooler receives and processes CREATE events from the spool. It
  parses each message and forwards it to the corresponding worker(s).

- When the worker is done, it notifies the master, which will delete
  the corresponding file from the spool.



## worker logic

- Each worker reads a configuration file containing the definition of
  the tags it can manage.

- Each tag is associated to a set of commands (URLs) and to a set of
  keyrings.

- A worker maintains a list of accepted scorsh tags, a map of
  keyrings[tags], and a map of commands[tags].

When a new scorsh message is received by a worker, it looks through
the commit history for commits containing schorsh-tags. For each
scorsh-tag found, the worker looks if the tag is among the supported
ones, then checks if the commit can be verified by one of the keyrings
associated to the tag. If the set of keyrings for that tag is empty,
the tag is not allowed to run (this is a quick way to disable tags).

(we might want to add an option to the definition of a scorsh-tag,
which allows to run the commands from unsigned and/or unverified
commits. This would be very dangerous though.)

Then, if the tag is allowed to run, the worker executes each of the
commands in the tag definition, replacing arguments as needed. If a
command is a script (file://...), then it must also correspon to the
hash specified in the config, otherwise it will not be executed.

When the worker is finished with all the commands for all the commits,
it will notify the master.