Why does Spark call "main" instead of using an interface like "SparkJob"?

Spark calls the main method on the specified class, and that main method is expected to construct a SparkContext or SparkSession. Configuration not specified explicitly, like master URL, will come from properties passed in on the command line or from configuration files.

What I'm wondering about: why does Spark look for a main method instead of using an interface like SparkJob with a run method that takes SparkSession as an injected argument? This is a common convention with other forms of managed code that run inside a container; you could still have a main method to construct and inject a SparkSession in local mode for unit testing.

In other words why not this

public class MySparkJob implements SparkJob {
  public void run(SparkSession spark) {
  }
}

instead of this:

public class MySparkJob {
  public static void main(String[] args) {
     SparkSession spark = ...
  }
}