Hey all - in watching the forum usage over the past few weeks I think there are a few improvements that we could make. This is a topic to give folks a chance to weigh in with their thoughts before we make these changes. If nobody opposes strongly to something then I’ll plan on implementing all or some of these changes in the next week. There are two major changes and I’ll break them down by section below:
Categories and tags
I think that our top-level categories have gotten a bit too numerous and in some cases too-specific for a top-level category. We should have as few top-level categories as possible, in order to maximize “signal to noise” (this is one of the first recommendations that Discourse gives), and each top-level category should have the same ballpark of activity (or at least, categories with almost no activity could probably be re-categorized as sub-categories until they grow enough).
In addition, we don’t make much use of tags in this Discourse, and it could be a way to deal with some of the categories that technical span multiple projects.
I’d like to make the following category/tag changes:
Add a general category at the top, merge Q&A and community chat into it
Move governance -> meta/governance
The notebook category:
Create a notebook/ipynb category for discussing ipynb files themselves, or the format
Create a notebook/classic-ui category
Move nteract to notebooks/nteract
Create tags for nteract categories and append to relevant topics
Create a kernels top-level category.
This would have discussion about creating kernels in the Jupyter ecosystem, and any projects that have a focus on kernels
Move Enterprise Gateway to kernels/Enterprise Gateway
Special topics
Move education to special-topics/education
Remove special-topics/security and make a security tag instead
Meta
Remove meta/announcements to just meta.
Make a release tag and add it to all release posts thus far
Move Governance to meta/governance and remove the “Public Steering Council” category, we can just use a tag for this
JupyterLab
Create a jupyterlab/extensions category
Badges
I’d like to add a few custom badges and to try and encourage people to notice and use badges more in general. In particular, I’d like to add the following badges and allow people to use them as titles.
Distinguished contributor
Team badges (e.g., jupyterhub-team)
Feedback?
Anybody have thoughts on this? In particular if anyone has experience with other Discourse forums and has thoughts, it would be much appreciated! (maybe @consideRatio has thoughts?)
Add a general category at the top, merge Q&A and community chat into it
Just a clarification: to move these to general/Q&A and general/community chat, or to make all three the same category: general?
Thanks for considering this! I think all ideas make sense.
Tag creation
Should we use a few predefined tags created by us as admins, or suggest any moderator to create tags following some guideline or common sense, or let anyone create them etc? And, for what groups? I think you can have tags valid on a group level etc rather than a global level.
I dont remember fully how their UI integration was, but am open to all these modes, but suggest to not start with allowing alls user right away until some experience has been gained.
I was imagining making them all the same “general” category, and adding a question tag for people that are asking questions. Maybe also a need-help tag to people with issues?
The reason for this is that I’ve found many people post tool-specific things in the Q&A section even if there is a more specific tool-focused section for it. Given how diverse the toolchain of Jupyter is, I think having this single mega-channels makes the signal-to-noise quite bad, so that’s why I’m thinking of reorganizing a bit.
re: tag creation - that makes sense to me. Should we just stick with all the current tags that we have? Or change them somewhat?
In general I am fine with the proposed re-organisation. Less (top level) categories is better I think. If the list is shorter we increase the chances of people reading all of the options and picking a suitable one.
The thinking behind the Q&A topic was to be a catch all “don’t know where else to put it” category. Aimed at people who are so lost they don’t know where their question should go.
Broader top-level categories gets my . It might sound counter intuitive to have “more posts per category” and “increase the signal-to-noise” but I think the key ingredient is “aim for categories with roughly similar amount of activity”. That will spread posts around and make the top level categories useful for filtering. Where as right now were (say) 90% of all posts go into one top level category, which means that level of categorisation is not very useful. I think focussing on the “key ingredient” will be key (haha…) and allow/require us to make top level categories which in a stricter “group things together” view might be made sub-categories. “Why is X a top level category?” - “Because it receives as much traffic as the other top level categories”. I think this would be a good “test” to use when deciding top level categories compared to doing it strictly by tool/community/language/etc.
Going slightly off into the wild: maybe the top level categories should be based around what people are doing, not tools. Looking at https://discourse.jupyter.org/latest a lot of things at the top seem to be in these two categories: “Help - I am trying to do something but it seems broken” and “Howto - I want to achieve X where do I start?”. Then have sub-categories (or tags?) inside these for specific tools.
This makes me think about who the categories are for. As a reader I hardly ever use them, in 99.9% of my visits I go to https://discourse.jupyter.org/latest and scroll through the list of what is new, judging if a thread is interesting or not from its title, author and who has replied. This means as a reader I hardly ever look at the categories.
When I post I have to pick a category and then I need to think about the categories. Not being able to pick categories we don’t have I then pick one that seems to fit best. Sometimes wondering if this is a JupyterHub or a repo2docker thing or maybe a general question/let’s discuss this. Often I feel confused/unsure what to put as a tag.
Could you make a list/layout of the final categories (and sub categories?) you are aiming for? That might be easier to read and think about than the current one which is more like a “diff” (so you need to know where we are now and then apply the diff, etc). In particular I want to know where BinderHub and JupyterHub will go.
Yeah - I was imagining that “general” could serve in a similar role. To me the challenge w/ these categories is that they are much harder to follow in strategic ways (thus the comment about SNR) and as a result are less-reliable for people getting the help they need, which feels lose-lose to me. I think large Discourse sites like ours only work if people can effectively limit which parts of the forum they pay attention to, otherwise it becomes a deluge of information that’s hard to manage.
I think that this is an interesting idea to think about, but probably something that we should explore over a longer period of time since it’d be a huge shift in the structure of the Discourse. How about we:
Define the kinds of “what I’m doing” categories that we’d imagine
Create tags for each of these
Find ways to encourage people to use more tags in general
See how those get used
Something like the following (here I am sorting based on category type and amount of activity each gets):
JupyterHub
JupyterLab
extensions
Notebook
classic-ui
nteract
nbconvert
Binder
Kernels
enterprise gateway
General
Special Topics (chris is wondering whether all of these should just be tags)
I haven’t seen any objections to the above proposal, so I’ll plan to start making iterative progress on this starting tomorrow unless I hear otherwise!
OK I believe that I’ve finished all of the category / tag reorganization, I put a next to the steps I took.
I haven’t done anything on badges etc. I’m not sure of the right categories to use. In addition I didn’t remove any tags, although it does seem that some are redundant with the categories (like jupyterhub or jupyterlab)
. It might sound counter intuitive to have “more posts per category” and “increase the signal-to-noise” but I think the key ingredient is “aim for categories with roughly similar amount of activity”. That will spread posts around and make the top level categories useful for filtering. Where as right now were (say) 90% of all posts go into one top level category, which means that level of categorisation is not very useful. I think focussing on the “key ingredient” will be key (haha…) and allow/require us to make top level categories which in a stricter “group things together” view might be made sub-categories. “Why is X a top level category?” - “Because it receives as much traffic as the other top level categories”. I think this would be a good “test” to use when deciding top level categories compared to doing it strictly by tool/community/language/etc.