Ansible Community Strategy 2023
(This was first posted on the Ansible GitHub Pages but is also posted here because I wrote it)
In the last article we covered metrics describing the community today, as well as some reasoning into why that might be. In this article, I’ll lay out my view of how to go about fixing it.
(If you’d prefer to consume this as a video, there is the recording from Ansible Community Summit in which I go over a lot of these points)
Communities as “neighbourhoods”
Let me start with an analogy, to try to illustrate what I think is missing in our community. For many people, the word “community” will conjure up an idea of a locality - perhaps a village or small town - which works to improve the place the people live in. This often looks like the following (at least where I live, in the UK :P)
- Most residents will agree that improving the community is worth doing - but they might not all agree on how. That’s mostly fine because people can specialise (you might organise maintenance of paths, etc; I’ll get funding for the new sports ground), so we can get lots done for the community as a whole. The mission is pretty clear, and it’ll only matter when there is opportunity cost of doing something.
- It’s clear that (in most countries, anyway) what goes on inside a house
is largely their business. So long as they’re not playing overly-loud music
or setting fire to their neighbour’s fence, it’s all good. Nobody wants to
police things to that degree (and it’s largely unworkable anyway).
- However, if one house starts to have an impact on those around it, then there are ways to sanction those people (legal options, yes, but also social contracts which are often stronger), and bring them back into line.
- It’s also true that while sometimes whole districts are planned at once, often development is more organic, as people buy land and build on it. A central planner isn’t needed (but is sometimes present).
- They will have a way to communicate with interested residents - crucially, even those who might not be active right now. This would have been a town notice board or similar in times past (and still is in my village, actually), but today it might equally be a messaging chat room or social website group.
- There will be a way to debate things of interest to the community - issues, upcoming events, local goverance elections, etc. Typically this is a town hall or similar venue where regular meetings are held.
That analogy actually works pretty well for us too. If we think of each of the projects and working groups as “houses”, forming a “neighbourhood” that is the whole Ansible community, then some conclusions naturally fall out it.
Firstly, we have a mission, and while it might be a good time to refine it, we are generally agreed that automation is our thing. Second, the “houses” (projects) are indeed largely independant, e.g DevTools does not tell AWX how to run their “house”. In fact, most of the projects are getting along just fine.
Where it gets interesting is the inter-project communication (between houses, if you will). Right now, it feels to me like the “social cost” of completely doing your own thing isn’t especially high, which leads to a wide array of behaviours. That’s fine inside a house/project but not when it leads to difficulties across the whole neighbourhod.
Relating to the lack of a central planner, this is fine - but all villages need power, water, connectivity, etc. This is usually the job of a regulator, and that distinction is worth exploring, probably in a later post.
However, it’s the last two points which are key to this post, as they directly relate to the two ways I think we can improve things in the Ansible community today - because a “notice board” is a website, and a “town hall” is a forum.
Websites, branding, and overloading
Let’s start with the “notice board” for our neighbourhood. As we’ve already
noted, there’s no single point-of-call for finding out what’s going on with
Ansible today. We have no way for the community to write about interesting
things, or announce new projects, etc. We have
ansible-announce, but the
mailing list does not get much traffic. We have the Bullhorn, but how widely
known is it?
Realistically, part of the problem is that the word “Ansible” means many things, and “ansible.com” can’t serve them all. It could mean:
- The language we write playbooks in
- The base package that delivers the language
- The full package with collections included
- The wider community / project
- The Red Hat product
The last two are the biggest problem. Today, ansible.com is Red Hat’s product site - that’s not a criticism, just a fact. However, as a result, the pages are product pages, the blog is a product blog, and so on. It would not be right to simply open this up to the community for editing, as it’s not the same Ansible we’re talking about.
The only way to resolve this is to separate the meanings. One could make the argument that the word “Ansible” belongs to the community, and the product should move to another site. That’s not wrong, but commercial systems move slowly, and I believe we need to fix this sooner. The lack of focus for new community members, and the lack of a voice for existing ones, is hurting us.
It’s common for upstream projects to have their own name, and web presence. So, I am proposing that we create a new short text (1-3 words) to reduce the overloading. We can debate it, but “Ansible Community” seems a good placeholder for now. We then use that to look at possible new DNS domains (ansible.community is available to us, for example), and then we can look at setting up a new site. I already have approval for this from within Red Hat, so we can start as soon as the community approves it.
In terms of function, at least to start with, I see a fairly standard setup - GitHub repo, static site generator, appropriate pages for contributor pathways, events, communication etc. Obviously a blog section too. This is all straightforward and standard practice, we just have to do it.
Of course, having the site be public and open to collaboration is key to improving those tenets I described last time. Also, I can’t foresee all the usecases we’ll have for this. The obvious thing to do is to create a new Working Group to oversee building it, and that will also be in my proposal.
- New short text for the community (to de-overload “Ansible”)
- New DNS property for hosting the community project
- New website
- New Working Group
Forums, fragmentation, and culture
Now to the longer, trickier part. I’m well aware that not everyone loves forums, so let me do a rare thing, and play the ad-hominem card. I really believe this will work - indeed, I have done this for multiple other communities and it has worked. If you find yourself unsure on this, give it a chance.
The biggest theme, by far, in all of these posts, in my discussions internally, in discussions with the community, everywhere I look, is fragmentation. Most of what we see before us today is at least partly attributable to the fact that we don’t have a common place to figure things out, or for new people to ask questions. We need to fix this.
For clarity, I am proposing to use Discourse as the software, as it is by far the best implementation I have seen. There is no “killer feature” that I can point to for why - it is simply full of hundreds of quality-of-life features, so many that I continue to be surprised by it even after 5+ years as a user. I think that it is no accident that it used by Python, Fedora, Mozilla, Ubuntu, Nextcloud, Pulp, LetsEncypt, Sailfish …. the list goes on. It has become the de-facto standard for async discusion.
Rather than try to list features, I’ll outline some (this is not exhaustive) of the things Discourse could help with, and then I’ll tackle some of the criticisms I think will come up.
A new generation of users, but existing users too
Nearly 2 years ago, I argued for including Matrix in our community (which the data shows to have been a good thing for our chat spaces). I did so by making the argument that we need to reach the next generation of users and contributors. That problem exists in more places than just chat.
Anecdotally, I have heard comments like “I wouldn’t know how to sign up to a mailing list if I wanted to”, “mailing lists are newsletters”, “sign-up means web, how do I web an email list?“, and so on. We need to meet people where they are today. Discourse supports basic email registration, but also login via GitHub, Google, Facebook, Discord, Twitter, and other SSO systems - at the least, we would have GitHub there.
Existing users are not forgotten either, as you can interact with Discourse by mail - other than updating your addressbook for the email to use for starting a new topic, little changes.
Meeting support needs
We’ve seen a migration to places like StackOverflow (SO), etc for support questions. We know that the #users:ansible.com (#ansible) room is busy. There is an unoffical Discord with many users, and multiple other Slack instances. There is no shortage of people needing help.
However, as SO shows, sometimes long-form slower replies are more useful than a chatroom. Chat rooms depend on the right people being online, and often question aren’t that urgent (and pastebins are less of a problem). I think the lack of a common community site/DNS has prevented us from creating this before, and so people went where they could (eg SO).
We can do a better job of helping our users, and building a dedicated support community around that. The use of the Solved plugin can give us similar behaviour to SO, but in a place where we get integration & signposting to the rest of the community & projects. Appropriate use of tags/groups can really help get the right people involved without overwhelming single people.
Things like support are needed by all community parts. Today knowledgable people who help other users are fragmented across too many places which dilutes that knowledge. A self-supporting user community is something I’d love to see more of, and the fact that I see exactly such a group of people (a) in other projects’ forums, and (b) in our own chatrooms, on SO, and in other places gives me hope.
Sync meetings exclude voices - both by timezone, and because quieter folks might not be comfortable speaking up. Also those with English as a 2nd language might struggle, both because the speed might make it hard to keep up, and because chat can be hard to use with a translation system. Being async-first means we get the best possible discussion, at the expense of speed - a worthwhile tradeoff, I think, as we tend not to move quickly anyway, and probably shouldn’t, if we want the full wisdom of the crowd (e.g. some people may be on holiday!).
One area we called out earlier was cross-project discussion, and I think this can hugely benefit from a forum. Having a place to discuss architecture, future plans, ways to solve issues, etc is necessary, but what really shines here is being able to include the right people, via groups & tags that actually work across the scope of the whole community (unlike on GitHub).
I’ve spoken with some people internally, and there is some support for bringing more of our internal architecture discussions upstream. This would be a perfect place to start such a thing, rather than creating yet another GitHub repo that no one can find.
Discoverability & archival
Mailing lists have a number of problems, but here I’ll focus on 2. Firstly, they are a high barrier - a user generally won’t sign up to a dev list, even though they might have a valid input for a given technical discussion. That’s a direct loss (see Wisdom of the Crowd in the previous article). Secondly, it’s hard to get people to sign up to new lists, and hard to decommission them later.
This translates into a particular problem with working groups. Spinning up a new working group has a high cost because people don’t know to sign up, and then when the WG is done, you have an old list hanging around.
Contrast this to a forum. New categories are free, and can be re-organised as well (and the URLs don’t break). So creating #working-groups/website is essentially free, and is immediately visibile to everyone who has a forum account (no new signup), and later could be moved to #working-groups/inactive/website later on. Indeed, if it becomes active again in the future, in can be moved back. Of course, new categories can be given incoming email addresses to function as an email list too. Much, much easier to maintain, and will have higher participation.
Categories can also come with their own rules, such as the use of up-voting or mark-as-solved, allowing each sub-community to act as it needs, while staying part of the cohesive whole.
Right now, I have an archive of 75k emails from
ansible-project on my disk.
The only reason I could get this out of Google Groups was because of personal
contacts - Google Takeout was broken (see
https://support.google.com/groups/thread/181378162?hl=en which has not been
answered). Google has demonstrated repeatedly that Groups is unmaintained (look
for example at the recurring issues with search, which have taken years to
fix, and some still aren’t), and I worry that we’d lose the archive of our
history, if/when Google decide to sunset it. This is very, very, very close
to vendor lock-in.
Similarly, we use Zodbot for our chat meetings. While I would like to see more of that go async (see above); for now, it happens, and the logs go to fedoraproject.org (although often copied to GitHub afterwards). Again, we don’t own that, and while I trust Fedora more than Google, it’s still not ideal.
All this can be imported to Discourse, and we can also import other things such as meeting logs and GitHub Discusions too. Just as I argued with Matrix, we should own our project, and having the new DNS gives us the chance to do so. I want to know that in 5 years time I’ll be able to go and read these kind of discussions on our community site.
In addition, things like better search, tagging, “similar posts” suggestions while writing, and so on make using that record easier for people too. But such things are powered by content, and thus work best if we bring our content together.
Groups have huge power in Discourse, on two levels. Public groups can be @-mentioned in a post and the folks in that group will get a notification. Groups can also be open to new joiners, or invite only, so for interest-based groups (eg. @edge) we can allow anyone to add themselves and be notified - but we can restrict access to others (eg @core) and make sure the right people are in there.
This mitigates the burnout / dogpile problem, where a helpful person answers users question, and then starts to get asked directly for help by more and more people. By tagging a group, they can be notified whilst we avoid naming people in a post, which avoids burnout. For people with high traffic, being in the right group(s) and subscribing to the right tags should keep traffic to the essential (and be easier to filter away from firehose of GitHub trafic, because it’s a different domain).
Trust & Decentralising Power
Discourse has several features that improve our ability to scale with minimal effort from the Community team, which speaks directly to the goal of decentralising power.
Trust levels are automatic (mostly) and almost any action can be tied to them. For example, to prevent spam a user must reach level 1 (takes about 10min) before being able to use email to reply to topics. But this goes much further, for example you could have a calendar category with all the community events (meetings, public demos, conferences) and anyone of (say) trust level 2+ could freely create events without gatekeeping. This has many applications for us.
Finally, private groups can be used to decentralise power. A group can have an incoming email address which bypasses the spam filters listed above. That means you can use a group to, say, register for hosting or social media, and the members of that group would have power over password resets, etc. Even the privacy, security, and code-of-conduct addresses could go here, and the membership of the groups can be updated as needed. Likewise groups can be used to democratise infrastructure support and so on.
Lots of small wins here, I think. First the downside - it’s true that mailing lists are naturally screen-reader friendly, being plain text, while a web interface can be less so. However, you can interact with Discourse by mail, and I believe that screen-reader compatibilty for the web view is decent these days (I don’t use one, but if anyone does, please test on meta.discourse.org as it will be up-to-date, and let me know).
On the plus, there’s full keyboard-only support, we can easily support other languages better (a category each perhaps? as discussed above, they’re easy to make), we can write modify the CSS/themes, and we’ll be providing a single place to access things that works - while a ML might work with a screen reader, does GitHub? Does ansible.com? Does trying to find the right place to join in?
Since I don’t need accessiblity I’m not sure what else is needed, but I’m happy to learn, please let me know.
Integration / Plugins
Modern forum software is web-based, which means it can integrate much more easily with other services. Links to GitHub, webhook integrations, single-sign-on, powering comments on blog(s) (yes, you can comment on the Bullhorn then), this list goes on. We can, for example, do things like managing signup for events through it, or integrating meeting agendas. I’m exploring using webhooks between Discourse and Matrix too.
Plugins are also a big feature, and can generally be enabled per-category. Popular things I would have from the outset are mark-solution, events & calendar, and possibly upvoting questions & translations. See https://meta.discourse.org/c/plugin/22 for more ideas.
Concerns on forums
I know a forum isn’t everyone’s favourite thing, and I anticipate some concerns. Many are valid, and part of the trade-off that I think is worthwhile, others I can argue against, so let’s have a look.
I like things the way they are
I can’t deny that this will shake things up - I can only hope that my data and argument have convinced you it’s worth the effort, that we have to shake things up.
I’m never going to tell people what tools to use (I didn’t with Matrix either), and we go to some lengths for compatibility with existing worklows (IRC bridges last time, Discourse supporting email workflows this time). But we have to also ask what’s right for the community as a whole, even if that means some changes for some of us.
Forums are old
They’ve certainly been around a while, and I don’t blame anyone who used a forum 10 years ago from being scarred by the experience - I was. But spend time on a Discourse site, and I hope you’ll be surprised. It’s no accident that so many projects are using it.
Forums are hard to maintain
Maintenance is a real concern, on two levels.
Firstly, the code/servers need to exist and be kept updated - but here we can draw on the fact that the Community Team has some budget. Much the same as when we funded the Matrix homeserver for Matrix, we can fund the forum. Paying another open-source company for their project feels right anyway, we should support each other.
More importantly, we’ll need moderators to run it. To start with this will be the people building it (i.e the Red Hat Ansible Community Team) by necessity, but we will be looking to build that group of people from the community. Such people already do good work in chat, on SO, and so on - but recognition/reward for that effort is hard to come by, and I think this can make it easier.
We should standardise on GitHub
This is a bit XKCD Standards isn’t it? GitHub is not a bad place for discussions, I agree - definitely better than a mailing list. However, I would counter this in a few ways.
Firstly, the fragmentation is worst on GitHub (470+ repos and counting). Do we pick a repo and bless that as special in some way? If we did, would it have rich enough tooling to cope with the scale of the project? That didn’t end well last time we tried it - we had to move everything out of ansible/ansible in 2020.
We also probably don’t want user support in that hypothetical special repo - reclaiming GitHub Issues for actual issues and having support in a dedicated place seems worthwile. And if a big chunk of the community is going to be where that support happens, don’t we want to make more use of that?
GitHub also doesn’t support @-groups across orgs, so unless we want to commit to a lot of overhead (duplicating groups), being able to involve the right people in discussions will be much harder this way.
Overall, the tooling around Discourse is much better for discussions, support, and groups - and it’s ours. We can, if we need it, develop plugins to provide additional functions - something that is never an option with an as-a-service platform.
It’s another place to check
I can’t mitigate this, it’s true. However, I would expect that the total amount of notification spam should go down (unless you choose to subscribe to everything, of course). Discourse can be used by web, mobile, or email, and you can set complex levels of filters to get what you need. Plus it comes from our own domain with decent headers so you can filter that way too. Oh, and you can set working hours so it won’t spam you at bad times - handy!
What happens to the mailing list / GitHub Discussions / other async places
Nothing, right away. I won’t sugar coat this though - if the goal is to reduce fragmentation, then they will need to be dealt with (archived & imported, which is entirely possible with Discourse). The timeline for that can be set as we go forwards, and each part of the community can move as fast as it is comfortable with. However I would like to see a critical-mass of the community adopt this fairly quickly - I’m looking for volunteers :)
- New forum for the community
- Hosts project-wide discussion
- Hosts support questions
- Other async discussion places get migrated over time
That’s a lot of text, so let me recap really quickly:
- The metrics are not good
- We have structural issues that are likely contributing to that
- We need to work on our fragmentation & lack of common voice
- We should do this with new website & forum
Hopefully I’ve convinced you that we need to own our virtual space, set up new things in it, and get as much of the community as possible to move to it soon. But perhaps you have questions? Of course you do, the crowd has wisdom.
Thus, accompanying these posts will be two discussion topics in GitHub (because we don’t have the forum yet!). These are:
- To create a new DNS & website, and a working group to build it
- To create a forum and look for parts of the project wantng to get onboard
Please do go and let me know your thoughts! I’ve deliberately not included rollout plans here (rest assured, drafts exist) because this is long enough. We can figure that out that out in our working groups once we’ve agreed the goals.
See you in the comments!