Because storing documentation in repos doesn't work great when you want to organize your documentation, discover or search it.
Having thousands of documentation files in a repo, next to the code is unmanageable, much more than thousands of documentation files in Confluence. In Confluence, you can put rights, tags, titles, organize in folders, assign owners, put comments, ....
Is Confluence good at it ? Not much, but it doesn't mean we should remove Confluence.
It's a problem.
But I still think that Confluence is better than nothing
Unfortunately the docs-in-repo + PRs-for-updates approach has way more friction for drive-by-corrections/commentary/Q&A from people not on the team, so it's not a holy grail either.
[0]: https://developer.atlassian.com/cloud/
[1]: https://gohugo.io
[2]: https://github.com/Redocly/redocI've taken to putting a link to the Confluence docs in the README so folks who find the code first can easily find the docs.
Middle ground I've found on some projects: very detailed code/data-oriented notes are in markdown in the repo, tied to a PR. Those doc files may reference external items like confluence pages or specific tracking ticket/URLs that relate to the code at hand.
I was on a team that had everything in confluence, and everything was impossible to find. The closest I came to understanding it was the confluence docs were always initial plans, but were rarely updated. When updated, you wouldn't necessarily know if you needed to look through 5 versions to see earlier thinking, or which links to 'updates' confluence pages you needed to trawl through. It was as much a problem of a growing set of contributors and growing departments than anything else, but there was a new 'direction' every 6-9 months (when new folks would come in) and "this worked at my old company" so they'd document stuff however they wanted.
No one on the dev team bothered to ever look there for anything, because it was simply pointless. Few people ever looked at it for anything more than "recent updates" to see what's changed in the last 2-3 weeks. Discoverability on the size of that project (and this is 'only' 5 years old ~80 people) was just useless.
A handful of folks did keep 'onboarding' stuff relatively up to date, but it was less than a year old at that point. I suspect that if those folks moved on, those docs may slowly rot.
On the whole, keep written docs both updated and useful and findable to a growing number of people with disparate needs and different contexts and backgrounds... it's a lot harder than it might seem when first considering it. Even if you have the people on a team with the aptitude for it, it's usually low priority in every work cycle, and the first casualty when trying to hit deadlines.
I once found a plug-in that did almost the same, allowed Confluence to read and render markdown files directly from (private) GitHub repos thus allowing me to get the best of both worlds!
- docs stays in the repo (thus is much more likely to actually get updated)
- docs get exposed in Confluence (thus is accessible for the business folks that does not do git)
- docs are easy to update (as you can use any editor: vi, emacs, IDE’s etc)
So conceptually the same as you’re solution ;-)
Unfortunately we did not implement the plug-in as we have too many eggs in the Jira basket (+1.000 users) which have the unfortunate side effect that the licensing price is derived from the +1000 users even though only the much lower number of devs would use it.
That is one repeated problem we face in the Atlassian stack (well a source of their income…)- even a seemingly cheap plugin (extension or whatever they’re called) ends up in the person suggesting how to ‘work smart/not hard’ having to find the funding typically a two digit thousands of € or year (thus I never bother with that anymore :(
Size of org/team is probably a factor, but the linking between the two products is one of the few things I see it has that most other tools don't. It's probably because most other tools are single-use, and they focus on one or the other, but not both sides.
And yes, it’s super bland and uninspiring. Just like Excel or Word. I consider it a feature.
Literally the only reason it exists. JIRA is the hook that gets companies on to the rest of the horrible Atlassian stack.