There are probably better ways to phrase this, but I couldn't think of any.

We need to figure out a format used to describe tasks to be run in taskbot. At the moment, everything is done with hard-coded executable names and arguments over stdin. This is very similar to how we did things in AutoQA and we already know there are problems and limitations with this.

The two options that I've been thinking about are to either keep the static executable names (ie, always have a consistently named executable for each phase) or to "compile" the actual job from YAML etc. kind of like how openstack's [[ http://ci.openstack.org/jjb.html | Jenkins Job Builder]] works.

Josef and I have talked about this a little and we're both very much leaning towards the YAML route due to it's flexibility to change things we need to change without requiring as many code changes in the tasks. Either way, this needs to be written up and sent out to qa-devel@