[yt-dev] Volume Rendering Refactor

[John ZuHone]

Hi Cameron,

Thanks for writing this--it's carefully written and lays out the issues well.

Just wanted to comment on this thought:

> The argument for the merge has been: "once it is in the main branch, people will sort out the bugs."

I don't think this description really catches the spirit of the argument correctly. As it is, it makes it sound like we are shoving the responsibility for finding and fixing bugs onto users.

For me, the "dev" branch is just that--development. The very fact we call it that carries with it a "caveat emptor." This is not to say that we just pull things into it willy-nilly: you have correctly emphasized that we should not accept PRs without appropriate review, sufficient documentation, and making sure the test suite passes. Ideally, this also includes adding new tests for the new functionality. I have definitely learned better coding habits from your efforts in this regard.

But the dev branch is by definition going to be unstable (which is why the alternative branch is named "stable"), even after all of these considerations. By getting code into the dev branch, we get a chance for users to take a whack at it in ways that most of us won't, which will by its very nature do a much better job at catching bugs in corner cases than programmed tests, which is the ultimate "test" for determining how "stable" this code is and whether it merits that label.

This points to the fact that we really do need to change things so that the stable branch is updated much more often than it has been, so that we don't have to keep shuttling people onto dev whenever they need a recent bugfix, which is frankly risky at the very least. That way we can consistently say "feel free to use the dev branch, but we have a separate branch called 'stable' for a reason."

That said, I think this particular PR definitely falls into a grey area for me as to whether it is ready for merging into dev or needs more review. Good arguments can be (and have been) made for both points of view. It sounds like Matt wants to do a final push with Sam in the Bay Area soon (or so I thought I saw on Slack), so maybe we should wait for that and then do the merge.

Best,

John