I'm watching <Readme.com>'s new editor preview. Lo...
# development
h
I'm watching Readme.com's new editor preview. Looks great, more like Notion. A user mentioned that they keep all of their docs in version control, then use Readme's API to sync. We already do this for reference docs. Thoughts on doing this for everything?
I think it would be hard to pull off right now because you cannot represent the widgets like callouts in Mark down, as I understand. But their new editor will make this possible I think
e
That's how we used to do things. At one point it was deemed wise to keep docs next to source - more likely to keep in sync in the PR that changes / removes / adds code. Then new wisdom said commits are hard and discourage doc fixes. So we've thrashed.
We used to have a 1st class - not repo local - confluence plugin to publish doc targets (markdown) to Atlassian Confluence even to help promote this.
b
in-repo docs empower non-maintainers to edit the docs, FWIW
h
Acknowledged. As someone who works on the docs a ton, my main issue with current setup is: • not having enough review of my changes • too hard to see the history of a page, e.g. git blame • too hard to audit when things need to be changed, such as when we rename a field. It would be very helpful to have ripgrep What I think maybe different from v1 docs: • we will still have distinct versions for each release • if you want to preview the change, you can copy and paste the code into readme. I vaguely remember it being too hard to know before how my change would render
b
(and allows docs to be changed atomically with code)
h
in-repo docs empower non-maintainers to edit the docs
They can do this via "suggest edits" button, which people often do use. Thoughts on which is more empowering?
e
in-repo docs empower non-maintainers to edit the docs, FWIW
This was nacked by the 2nd wisdom phase. Some consensus said commits were hard and discouraged non-maintainer contribution to docs - notably fixes.
c
+1 for in-repo docs. Makes it much easier to keep up-to-date with code changes, as they happen, imo. 🙂
👍 2
👎 1
e
I am a 1st wisdom phase guy myself.
h
and allows docs to be changed atomically with code
This would be great for making the docs audit I do every major release less painful. It usually ends up taking me 3-4 days to audit 🤷
e
I would definitely encourage, as always, a wisdom thrash post-mortem if we switch back.
☝️ 1
b
I think instead of saying "commits are hard" we should find out why and solve that problem. E.g. going through the test suite is annyoing, perhaps we can find out how to skip that, etc...
h
I would definitely encourage, as always, a wisdom thrash post-mortem if we switch back.
I am probably rewriting my recollection of events, but I remember the number one reason we changed to read me is that we needed versioning of the docs. The number two reason is that we wanted the docs to look prettier
E.g. going through the test suite is annyoing, perhaps we can find out how to skip that, etc...
Yeah we skipped CI for docs only changes
e
but I remember the number one reason we changed to read me is that we needed versioning of the docs. The number two reason is that we wanted the docs to look prettier
You can clearly do both and have the primary docs source come from the repo as is being discussed now and as it was in the 1st phase of wisdom with the confluence plugin / doc targets.
So maybe it wasn't wisdom switch but expediency.
1
The big problem here is can you turn off edits in Readme? Because how would we sync those back?
IIRC for confluence we couldn't so we had header / footer to warn you cannot edit. Fairly lame.
h
The big problem here is can you turn off edits in Readme?
This has not been an issue for the generated reference docs. The few people with edit access to read me.com are trained to not edit directly. When users suggest an edit, we comment on their suggested change redirecting them to make a pull request
My main open question is how difficult it will be to write read me.com compliant mark down. For example, what will it look like to create a call out Block. I think this might be changing soon with their new editor, which I will get better access to in the next two weeks
h
I was strongly against docs-in-repo, mostly because that led to them not being updated enough. Being able to edit ~inline as you read them in readme is powerful.
So I am generally not a fan of going backwards there.
Definitely not a fan of having to edit source that has to go through laborious process (PR, review, merge) for any change, and that only later gets generated into docs.
Our switch to readme.com is one of the reasons our docs are dramatically better than they have ever been
e
To be fair, without hard numbers, besides looking better, which is all due to Readme.com, I think the reason the docs are better in >80% territory, is because of Eric. Maybe I have a wrong idea though of where all that content has come from.
1
h
There are good theoretical arguments in favor of in-repo docs, but I have not seen them play out well in practice. Ease of modification seems to trump all other considerations in practice.
b
FWIW I would just be using the GitHub web UI for edits, so it's still relatively easy (to me) If anything id feel more empowered knowing my changes will be reviewed. I don't have high confidence in my prose.
a
ReadTheDocs (i.e. hosted Sphinx) covers a lot of the ease of modification difficulties that other in-repo docs solutions have, mostly because it’s a continuous deployment platform
i.e. once a commit lands on a given release branch (or
main
) , the relevant docs are updated
very much agreed that in-repo docs are useless without CI/CD of those docs
h
Oh. That's cool, they are adding a new feature for collaboration kind of like google docs and git. That address is my biggest frustration: how hard it is to get reviews
b
Another bullet for in-repo docs, we can format and lint them. E.g. find stale imports, use good code style, etc...
👍 1
h