Direct user help questions to Stackoverflow

[dashohoxha]

This is way more effective (and searchable) than mailing lists and chat rooms.

For example, when I have a problem, my first reaction is to google it, and more often than not I find that someone else has asked the same question on stackoverflow (or stackexchange). Sometimes the answer comes from an issue (for example GitHub issue), and more rarely from a discussion thread in a mailing-list archive.

Asking a question on a mailing list needs to first find where is the mailing list, and then subscribe to it. After asking your question (and hopefully getting an answer) you have to unsubscribe again, because the pile of unrelated and uninteresting messages becomes unbearable. Too much overhead for asking a simple question and getting a simple answer.

Asking a question on a chat has almost never worked for me. You have to hope that someone who may know the answer is online at the same time as you. I think that chat may be suitable for people who already have some familiarity with each-other, not for asking random questions to random people.

Some stakoverflow forums where we can direct the users:

• https://stackoverflow.com/questions/tagged/numpy
• https://datascience.stackexchange.com/questions/tagged/numpy

We should promote somehow these links on the website and encourage users to use them.

[InessaPawson]

There is a NumPy group with 636 members on Reddit as well. reddit.com/r/Numpy/ It looks like it is fairly active.

[stefanv]

For reference, this is the solution the imaging community uses: https://forum.image.sc
Jupyter does something similar: https://discourse.jupyter.org/

[rgommers]

Discourse is more a replacement for mailing lists than for Stack Overflow. Stack Overflow works really well, other things in the past haven't. So I see no rationale for pointing people anywhere else.

[stefanv]

I had the same feeling initially (and, personally, I am no great fan of Discourse), but I do think it establishes a better sense of community than StackOverflow, which can be quite harsh and isolating. But, sure, I have no horses in this race. Maybe @choldgraf has thoughts too.

[choldgraf]

hehe I do have thoughts. Just to quickly give two points:

• I actually do think of Discourse as a replacement for something like Stack Overflow, in the sense that it is a place where people can ask questions and get help. In fact, one of the creators of Discourse was also one of the founders of stack overflow. Discourse communities are basically just online forums but where you (as the maintainer of that community) have way more control over the kinds of conversations, guard rails, etc that take place.
• Not to pick on Stack Overflow, but it has the potential to be a fairly toxic place, especially for newer or more inexperienced contributors. I personally wouldn't feel comfortable relying on it as a "recommended" means of communication in a project because conversations on SO often don't reflect the values of the projects I work on.

[rgommers]

Not to pick on Stack Overflow, but it has the potential to be a fairly toxic place, especially for newer or more inexperienced contributors

More so than a mailing list, or IRC, or ....? I don't think that's true, I think any kind of communication mechanism like that can have similar issues. And certainly our numpy-discussion mailing list does.

Anyway, that's not really the issue here. The thing is: Stack Overflow simply works way better than anything else we've tried (and not just for NumPy). People seem to be motivated by the SO points system and interface to answer questions. The majority of those people will not answer questions on a mailing list or forum that's specific to the project. Which means more unanswered questions (already common on the mailing list), and a bigger load on maintainers.

On top of that there's the burden of maintaining Discourse. That's also a nontrivial amount of work to set up and keep going.

[rgommers]

I also checked activity: there were about 14,000 questions with the numpy tag in the last year. There's simply no way we can deal with such volumes.

I appreciate the suggestions (Reddit also), but anything but Stack Overflow seems like redirecting people to the wrong place.

[stefanv]

Edit: It was pointed out to me that the post I linked to below is rather aggressive and ugly itself. I agree, and I was not trying to condone that sort of response. It just seems to me like there is a lot of unhappiness with the vibes on S/O, and that post was a way to surface how bad it makes some people feel.

I would be bit careful with "S/O simply works way better than anything else we've tried (and not just for NumPy)". We have not tried very much or hard, and I'm not sure which other projects you're referring to. But I thought the whole idea of the redesign is to try and do things better than ourselves in the past.

I understand your points w.r.t. volume of questions and motivation for those answering: but perhaps that only satisfies the requirement of the 95% out there we are already serving. We are specifically interested in expanding our reach to new developers who may not feel as comfortable in the harsh S/O landscape.

You read a post like this and all sorts of things may run through your head. But when you click through and see what especially women write on Twitter and elsewhere, it is clear that S/O is not a place in which they feel comfortable. Unsurprisingly, it is not my personal experience, but I cannot deny theirs.

One of the reasons we set up the Slack forum was because we wanted a safer place for beginners to start, and I think we had the right idea there.

I cannot volunteer to operate a Discourse forum, so all I'm doing here is to ask that if we accept S/O as our default help forum we keep in mind that we may be alienating some beginners—with an eye toward making a better arrangement in the future, maybe with input from some of the community organizers we heard at SciPy2019.

[matthew-brett]

I wonder if there is any way we can moderate the SO numpy questions, to encourage gentle responses to beginner questions? Maybe SO would be interested in a discussion on how best to make that work?

[stefanv]

There seems to be some interest: https://stackoverflow.blog/2018/04/26/stack-overflow-isnt-very-welcoming-its-time-for-that-to-change/

For scikit-image, we have several people subscribed to the relevant tags who monitor responses.

[rgommers]

We are specifically interested in expanding our reach to new developers who may not feel as comfortable in the harsh S/O landscape.

That's a fair point.

I cannot volunteer to operate a Discourse forum

This really is the main issue. I'd be all for multiple sites that cater to users with different needs, but I also cannot volunteer for this. We could turn this around: let's encourage people interested in this to work on it, if it gets traction then we can make it "official" (provided those people continue to volunteer to maintain it). For starters, it can be the same as things like Gitter/IRC/Reddit/etc, which also have smaller groups of people that get something out of it. It just cannot be official for now, because that's a major negative on maintainer load (and keep in mind that maintainer burnout is one of the main issues in open source).

We have not tried very much or hard, and I'm not sure which other projects you're referring to.

NumFOCUS had a Discourse, and Zulip, and some other tools/sites I can't even remember. All lots of effort, and all turned into dead zones.

For NumPy (combined with SciPy in some cases), we have or had multiple mailing lists, IRC, Gitter, Slack, StackOverflow, Reddit, wikis, a doc editor wiki, Planet SciPy, SciPyCentral. I might still be missing something. I'm also not sure it's visible to you or most others how much work it was and is to deal with a mailing list migration, keeping a server alive, etc.

I have no strong opinion on the merits of Discourse from a user perspective. However at this point in time something like Discourse, or anything else really that requires running a server for the project (or paying a significant monthly sum for a hosted version), is a complete nonstarter simply based on the negative effects on maintainer load.

so all I'm doing here is to ask that if we accept S/O as our default help forum we keep in mind that we may be alienating some beginners—with an eye toward making a better arrangement in the future, maybe with input from some of the community organizers we heard at SciPy2019.

Why not tweak the text a little to address this issue? All I want is a text that reflects the reality of where people can go to get their questions answered. Something like:
User questions are best asked on a public community site. The largest such site is Stack Overflow (link). There are also smaller active groups of users that ask and answer questions on IRC, Gitter, and Reddit (links).

[stefanv]

We are specifically interested in expanding our reach to new developers who may not > NumFOCUS had a Discourse, and Zulip, and some other tools/sites I can't even remember. All lots of effort, and all turned into dead zones.

Ah, sad, I like Zulip so much more than Slack.

For NumPy (combined with SciPy in some cases), we have or had multiple mailing lists, IRC, Gitter, Slack, StackOverflow, Reddit, wikis, a doc editor wiki, Planet SciPy, SciPyCentral. I might still be missing something. I'm also not sure it's visible to you or most others how much work it was and is to deal with a mailing list migration, keeping a server alive, etc.

All too visible, unfortunately :)

I have no strong opinion on the merits of Discourse from a user perspective. However at this point in time something like Discourse, or anything else really that requires running a server for the project (or paying a significant monthly sum for a hosted version), is a complete nonstarter simply based on the negative effects on maintainer load.

Agreed—that is the reason why I haven't even tried it for any other projects.

Why not tweak the text a little to address this issue? All I want is a text that reflects the reality of where people can go to get their questions answered. Something like:
User questions are best asked on a public community site. The largest such site is Stack Overflow (link). There are also smaller active groups of users that ask and answer questions on IRC, Gitter, and Reddit (links).

I like that text a lot; thanks, Ralf.

[InessaPawson]

Stack Overflow is a good source of information on using NumPy without a doubt. However, as other participants of this discussion pointed out, some SO moderators can be very curt and condescending in communication, which doesn’t reflect the culture and communication style of the NumPy team.
While we are figuring out a friendlier communication channel for the NumPy community, let’s put a disclaimer next to 'Stack Overflow' on numpy.org 'Community' page stating, 'This forum is not moderated by the NumPy team'. Let’s add the same disclaimer next to 'Reddit', because we are not trying to pick on SO.

[rgommers]

Done in gh-178, added notes on no moderation and requests to respect the CoC.