On Thu, Jul 23, 2015 at 07:31PM, Ted Dunning wrote: > Branko, > > Let me preface this by my thought that I feel that the JIRA's are a > valuable organizational tool in that I should be able to find the > information on a particular issue there. The Hadoop or Zookeeper JIRA's, > for instance, are often models of this. Issues are discussed, reviewed, > dissected and finally a resolution is made. > > More comments in-line in a rather ruthlessly summarized quote: > > > On Thu, Jul 23, 2015 at 7:11 PM, Branko Čibej <br...@apache.org> wrote: > > > > > >> > > >> The important question is whether the development process is open, not > > >> whether some entries in Jira appear to have adequate comments. > > > But, from what I can read in the comments about it, and from what I can > > > see when I scan the tickets, lists, commits etc; The commits only refer > > > to JIRA tickets and not discussions on the dev list, the JIRA tickets do > > > not refer to anything, and the dev list does not refer to neither the > > > commits IDs nor the JIRAs...so how exactly are we to interpret what's > > > going on then, if it's all suddenly irrelevant? > > > > > > Open Source development is not just about publishing your code, it's > > > also about making the development and decision process open and > > > transparent, and in several cases, such as the ones Ted listed, it does > > > not appear to be that way yet. > > > > > > I see that this issue has been acknowledged on the dev list by at least > > > one member of the project, and while that is a positive response, I > > > stand by my decision to withhold support for graduation till I am > > > satisfied that this has been shown in a consistent manner across (most > > > of) the board. > > > > There's a bit of an impedance mismatch here, I agree. I insist that Jira > > is not relevant history. > > > You may or may not claim that, but the fact that issue tracking is required > to be on Apache controlled resources indicates a somewhat different > result. Also, JIRA is often used as a way of putting comments onto the > mailing list, while at the same time organizing them for ease in finding > them. > > So JIRA may or may not be relevant history (and it certainly is relevant in > many ASF projects), the fact is that I only used it as one source of > information. I would be happy in finding discussions in any of the > standard places, but I find that many issues seem to have essentially no > discussion. There are comments about reviews that are done, but there is > no record of those reviews. > > > > > > Discussions do happen on the dev@ list, so the > > problem must be in the commit messages. > > > Hmm... I think I disagree. > > It is clear that some discussions happen on the mailing list, but it is > also pretty clear from what I have seen so far that many discussions don't > happen on the list. I can only judge by what I can see and from my > examination of recently closed issues, I found no discussion of these > issues on the dev mailing list. > > The fact that *some* issues are discussed does not lead to the conclusion > that all design discussions are occur on or reflected onto the list. The > fact that a random-ish sample of recent issues reveals no substantive > discussion even though some of these issues seem to be worthy of discussion > (race conditions and such) makes me think that there might well be > substantial discussion going on somewhere else. > > What would make me happy is a comment on the JIRA which is then reflected > on the mailing list of the form "Joe and I talked about this and it looks > pretty simple" or "We talked about this as a group and we had the following > suggestions which were all hooey so we are doing this". > > I don't see that happening. I see issue opened. Issue closed. > Occasionally a "reviewed by" comment. All that is only on the JIRA and > nothing appears on the mailing list. > > > > ... Discussions happen on the > > dev@ list but a Jira issue is raised for each every change, ever so > > minor; the notification about the issue creation goes to the dev@ list, > > the change is made, nobody objects and that's it. Hence, there doesn't > > seem to be much correlation with all the JIra spam and dev@ discussions. > > > > Could be. > > > > First of all, it's not reasonable to expect a dev@ discussion for every > > one-liner change; CTR rules. Next, it's not reasonable to open a Jira > > issue for every one-liner change; that's simply a waste of time (and > > leads to the kind of misunderstandings that we have on this thread). > > > > I agree. But the changes I am seeing are not really one-liners. Or at > least they are not exclusively one-liners. > > > > I do insist that discussion of important issues and features does happen > > on the dev@ list. The Jira tickets that are created as a result of those > > discussions can easily be cross-referenced by a simple search in the > > dev@ archives. > > > > I have tried to find such. > > Can you provide some examples? I keep looking and citing the examples that > I find, but none of them show the behavior that I would like to see. > > > > > My only recommendation here would be to use Jira only to track important > > issues and to always write proper commit logs. The latter is an art that > > takes years to learn ... > > > > We seem to agree quite a lot in this. I think I have a lower tolerance and > would like to see a JIRA on most changes, possibly annotated with > "one-liner". Or a git message to this effect.
Thank you for the thoughtful exchange, gentlemen! JIRA or not - I agree that decision making for things that matter can be improved. Here's one more background point to help you to see the situation in a slightly different light: for most of the original team English might or mightn't a communication tool. You might find it hard to believe, but there are still places in the world where software developers do not use English to communicate design ideas to each other ;) Hence, the discipline of meticulously writing down every verbal comment verbally (in the mother tongue) takes time to be developed. And as has been pointed out there's a huge improvement in this area that the community has already undertaken. And it continues to get better literally by the day. As has been said a number of times: the incubation is about teaching a new community what it means to follow the Apache Way. It's not about hand-holding and training in the JIRA, Bugzilla, or git. Did the community learn what it means to develop the software in the open and follow the spirit of the Foundation? As someone who was with the project for the long time I think the answer is yes. At the same time I do recognize that there are things that could be done better. But can anyone point me to an ideal Apache project? Hadoop has been cites as an example of JIRA handling, but if you look around you'll find tickets similar to HADOOP-12235, which has one human comment "+1 lgtm" and a bunch of CI integration comments. Will we be reaching the graduation consensus today? It isn't even most important. What is most important is if the IPMC wants to take on a role of technical and development process arbiter and make all podlings to follow the same set of guidelines, which still aren't clearly indicated as 'must to follow' nor agreed on by the said body of the people. Cos
signature.asc
Description: Digital signature
