I'm trying to understand the changes in https://issues.apache.org/jira/browse/MESOS-1453 and the SchedulerDriver JavaDoc.
In the 0.19 behavior, it made sense to me that a framework would hold onto a copy of all the latest task statuses it knew about, and could poll reconcileTasks with these statuses in order to request delivery of any lost messages (covering the case of both the framework being absent for a while or just a general loss of messages.) Is the idea behind the changes in 0.20 that a framework now need only call reconcileTasks once after registering with a master? In that case, what is the use case for having the API still take a list of taskStatus objects - so frameworks can decide that they don't want to know about unknown tasks [1379]? If frameworks should still routinely ask for missing messages - then why bother sending all updates and causing the framework to have to handle the work of routinely ignoring duplicate status updates? Thanks, -Whitney
