Reviewed HDFSEventSink with respect to "escape sequence" and RollingFileSink.
Here are some of the things I found:
1. Looks like we will be able to reuse BucketPath, which is good.
2. HDFSEventSink has a Object called BucketWriter. One of the major things this BucketWriter does is manage the rollover lifecycle of the file.
3. RollingFileSink handles all the rollover lifecycle it self.
1. To complete 1295 the RollingFileSink will need to have something link a BucketWriter.
2. I would hate to make two bucket writers. So I will connect with the Flume brain trust and see if we can build a bucketWriter that can be extended to handle HDFS and normal filesystems.
Let me know what you think