I am saddened to have to start this thread *again*. While I thought we had reached consensus on this, several times over, apparently some people disagree. I hope this will be the last time.
With this thread, I am asking the community to reach consensus (1) That 208 should be merged this week, without further delay; and (2) That Moon Lee Soo and Felix Cheung take no further part in the discussions of 208 and 702. This PR has been pending since August. It has been stalled that entire time for no technical reason. We reached agreement to merge 208 in November, again in December, and again in February -- when Moon agreed to stay out of the issue. At that point, Alex, I, and others, began working on it, and appeared to be making substantial progress. And then Alex just stopped. Instead, he commenced the thread saying that a consensus had to be reached on 208 and 702. Until that point, essentially no-one had paid attention to 702. In the discussion that followed, we reached a consensus to merge 208 as soon as possible. After the thread had died, Alex asked if anyone had additional comments, and Moon popped-in to insist that both PRs be merged. Again, no-one supported 702. At all. Each time I said "we had a consensus before, does anyone want to change it," Alex or Moon steered the discussion away. The final vote was not to merge 702 or merge "both" -- it was to treat them as normal PRs. (Although one person did want both merged simultaneously.) That would mean completing 208 on its merits and then evaluating 702. At the time, I objected to the discussion, because I thought the whole thing was a contrived excuse for Moon to reject 208 by pushing 702. That is exactly what he is now seeking to do. *Status of 208 & 702* PR 208 has been feature-complete and testable since early September. It has been adopted by more than 1000 users, who I have been supporting for more than six months. The code has not undergone any major changes since September. There are no known bugs, and no outstanding feature requests that can be satisfied without major changes to the Zeppelin architecture. 208 does *not* fail CI. 208 includes extensive unit tests of the R-Spark integration because this turned out to get broken by changes in Zeppelin often. Because CI is unable at present to provide a consistent environment, 208's *OWN UNIT TESTS*, which pass when run on an ordinary machine, fail when run on CI. 208 does need a push for compatibility with a recently adopted PR -- that is work I've essentially completed, but have not pushed. PR 702 is a re-design based on 208 -- not just architecture, but right down to the choice of demo images, which were taken from 208's documentation. In fact, 702 has had been re-engineered several times to more closely conform to 208's architecture and feature set. But 702 still remains feature-incomplete -- it cannot handle the range of visualizations, R classes, etc., that 208 can. It is not stable code, and shows no signs of stabilizing any time soon. No-one has adopted 702. It has changed radically, fundamentally, at least 4 times over the past two months since it was submitted. One of those changes was only days ago. 702 also has no proper tests, which is the excuse for not merging 208. 702 has things labelled "tests," but they don't actually attempt to connect to R or Spark, which are the things that break and which therefore need testing. *** I would like credit for my own work and design. I think I have more than earned that.
