Hi everyone, I'll be chiming in from time to time as Anthony Scopatz and I help several people from Quansight become more integrated to community Arrow development. I'm a fan of Arrow's goals and have similar goals for a cousin project called http://xnd.io. I'm eager to find ways to collaborate on the compute infrastructure between the two projects, in particular, for example.
Here is my $0.02 on this issue, Chat rooms can be a useful mechanism for engaging with new developers. However, Slack itself does not really allow for the kind of large-scale community participation that Gitter allows for. If you have a chat room I recommend Gitter. All that said, I would personally favor a discourse (https://www.discourse.org/) solution over chat rooms. I've noticed several younger folks not really liking the mailing lists and seeking out chat rooms first --- the success of pytorch, mxnet communities indicate that perhaps they could be encouraged to use something like discourse. On a related note, could someone help me understand the relationship between Github Issues and JIRA issues? Is one preferred? I understand that contributions to the code are recommended as PRs on Github. Does that mean a branch at Github is considered to be the primary repository, or is there another place that code has to go to be official? I suspect this is all documented somewhere. I would welcome a simple link to the right place to read. Thank you, -Travis --- Travis Oliphant Quansight On Thu, Jun 21, 2018 at 3:25 AM, Wes McKinney <wesmck...@gmail.com> wrote: > hi all, > > I wanted to bring up some concerns I have about the Slack room hosted > at http://apachearrow.slack.com. > > Corporate communications have changed a lot in recent years with the > new wave of IRC-like chat systems such as HipChat and Slack. In many > companies, Slack has become a preferred form of communication over > e-mail or other asynchronous messaging tools. This trend is negatively > impacting Apache Arrow in some ways that I will explain. > > Initially we created the Arrow Slack channel as a means of secondary > communication, to facilitate real-time discussions and help build the > community. So people, particularly newcomers, are coming to the > project and seeing 4 ways to communicate: > > * dev@ Mailing list > * JIRA > * GitHub > * Slack > > As a result of broader trends in the world, they are electing to use > Slack as their first, primary channel to interact with the project. > This is bad for many reasons: > > * Slack is essentially private. While anyone can join Slack, chats are > not archived in any public place, nor are they searchable through > internet search portals. I do not think it meets the public > communication requirements of Apache projects in general > * We've exceeded the message limit for free Slack channels; upgrading > to a paid Slack plan for Apache Arrow, with 650+ members, would be > very expensive > * Only 3 out of the top 20 Arrow contributors (by # of commits) are > regularly on the Slack channel. I don't use Slack, for example, and I > would rather not be expected to > * We are geo-distributed in many time zones; even if we all used > Slack, synchronous/real-time chat to discuss the project is frequently > impractical > > Because of the "real-time" nature of IRC-like systems, people's > discussions and questions get intermingled, so keeping track of > longer-running discussions may be difficult. It's hard to know when > someone's question has been answered or whether people have > sufficiently discussed a particular topic. > > Many discussions or questions are by their nature asynchronous, and it > may take 24-72 hours or more for Arrow contributors to make a > thoughtful reply. > > As a result of all of this, we are missing opportunities to have > deeper discussions, develop the Arrow roadmap, create new JIRAs to > capture bug reports or feature requests, and other activities of > healthy open source communities. Additionally, the private nature of > Slack is causing organizational knowledge (particularly Q&A / FAQs) to > essentially be lost. Users with questions won't stumble on answers by > searching on Google (as they would with a mailing list or > StackOverflow). > > I don't think Slack is necessarily bad for users in a corporate > environment; in many companies it is expected that all people will > have the Slack client open at all times. This isn't the case here, > though. > > My strong preference in light of the activity I have been observing on > Slack (which I encourage you to explore yourselves) would be to close > the channel and direct discussions or questions take place on the > mailing list, JIRA, or GitHub (all of which are archived on one or > more ASF mailing lists). Since migrating to Gitbox, we have enabled > GitHub issues on the repository, which has helped lower the barrier > for newcomers, but a large percentage of the time GitHub issues would > be better as JIRA issues or e-mails (which is what the GitHub issue > template says, alas). > > Interested to hear the thoughts of others on this. > > Thanks, > Wes