Interesting question posed in this JIRA: Thoughts? The basic thrust is:
Should bigtop implement tests which it knows are likely to be out of date,
or inferior to tests which we can actually write better versions of, just
because they are provided by the underlying framework?
Also - should bigtop tests use the mapred.* APIs?
I guess its a fundamental question that gets to the heart of what bigtop
should do: i.e. , should we maintain our own tests when it is in the
interest of bigtop users? or should we always proxy to external ecosystem
components for smoke tests whenever possible, even if we know those tests
are less than idea.?