Be aware that job-2 depends on job-1 being started. The started state of a script is reached before it is finished, so it is actually possible for the messages to appear in the logs in random order. But unless the system is very heavily loaded, there will only be split-second differences between the moments they run. A safer way for job-2 to depend on job-1 is by depending on a signal emitted by job-1 after it has obtained its goal. In the case of our rationale, etc/init/mountall-net.conf will emit mounted for filesystems that have been mounted. That would be a good signal to depend on:

start on (starting kdm and mounted MOUNTPOINT=/home)
	  

Having job-3 depend on job-1 through job-2 of course means that job-3 will never run if nobody starts job-1. This can be remedied by having job-2 trigger job-1.