I don't think anyone ever regretted taking some extra time to test and verify something critical before deploying it.
Thanks to the QA Hackathon I have gotten a lot of extra eyes and review to look at what I have been working on. Many of you have probably noticed a recent upswing in code churn and discussion related to this. Because of this I have decided that more time is needed before the new stuff is deployed as stable. I do not expect many people to complain about taking some extra time to make sure things are right. I am not abandoning the plans, efforts, or decisions reached at the QAH. I am however extending the timeline, and will insert an extra step between now and stable-release. The main issue here is that the QAH focused mainly on backwards compatibility and testing that nothing is broken. This happened because it is BY FAR the most critical thing. The QAH discussion ended with a punch-list of things that needed to be done before we can replace the internals of Test-Simple. I will stick with this punch list, and before anything goes stable it will be done, with rjbs, xdg, and the others involved to verify it. However this will happen a bit later than we expected. What new eyes, and attempts to use the new stuff (Test-Stream) have revealed is that there are still some design elements in the new API that need work. These are mainly in things that are completely new, and do not have backwards compatibility to drive them. Once there is a stable release these become hard to change. The fact that I have found the need to make 2 significant API changes in the weeks since the hackathon tells me that it would be irresponsible to release it right now. Others have also expressed concern after seeing the recent churn. The recent discoveries and improvements have largely been driven by the new interest and usage of Test-Stream. If usage tapers off then the pipeline likely will as well. Because of this I do feel there is a need to make Test-Stream, the new stuff, available as an experimental and independent release fairly soon. This dist will be available for early adopters and interested parties to play with and provide feedback. Existing tools, specially toolchain level ones, should NOT switch whole-sale to Test-Stream while it is in this experimental phase. I am also trying to get a handful of people willing to commit to helping me with review, I don't like working in a box, and I suspect others don't either. So to be clear: * Not forking the ecosystem, the Test-More/Builder/Simple internals will be changed, but not yet * Test-Stream will have an experimental phase where it will be released to cpan for people to play with * The punch-list, and the QAH discussion are still key to any future stable release * I am just taking a bit more time to make sure I don't lock people into another insufficient system. -Chad 'Exodist' Granum
