I've found that a bare repo over SSH is the simplest way to keep control and reduce attack surface, especially when you don't need fancy PR workflows. I ran many projects with git init --bare on a Debian VPS, controlled access with authorized_keys and git-shell, and wrote a post-receive hook that runs docker-compose pull and systemctl restart so pushes actually deploy. The tradeoff is you lose built-in PRs, issue tracking, and easy third party CI, so either add gitolite or Gitea for access and a simple web UI, or accept writing hooks, backups, receive.denyNonFastForwards, and scheduled git gc to avoid surprises at 2AM.
I'm going to trust the constant stream of updates from the company itself which shows exactly what went down and came back up rather than a random anecdote.
Recent years have shown this to be the wrong prediction strategy. The reason seems to be an incentive imbalance where there are quite a few reasons for companies to lie (including their own CLAs) and not a lot of repercussions for doing so (everybody competes on lock-in, not on product). Of course, the word-of-mouth approach is also exploitable by dishonest actors, but thus far there doesn’t look to be a lot of exploitation going on, likely because there’s little reason to bother (once again, lock-in is king).
This seems intelligent, after all companies are incapable of making errors in reporting and also have absolutely no incentive to lie about stuff like that. Those 500 errors others have reported as experiencing must have just been the wind.
I rarely successfully get Codeberg URLs to load. Which is sad because I actually would very much like to recommend it but I find it unreliable as a source.
That being said, GitHub is Microsoft now, known for that Microsoft 360 uptime.
I have never had this issue. IIRC Codeberg has a matrix community, they are a non-profit and they would absolutely love to hear your feedback of them. I hope that you can find their matrix community and join it and talk with them
These days it feels like people have simply forgotten that you could also just have a bare repository on a VPS and use it over ssh.
Most developers don’t even know git and GitHub are different things…
I've found that a bare repo over SSH is the simplest way to keep control and reduce attack surface, especially when you don't need fancy PR workflows. I ran many projects with git init --bare on a Debian VPS, controlled access with authorized_keys and git-shell, and wrote a post-receive hook that runs docker-compose pull and systemctl restart so pushes actually deploy. The tradeoff is you lose built-in PRs, issue tracking, and easy third party CI, so either add gitolite or Gitea for access and a simple web UI, or accept writing hooks, backups, receive.denyNonFastForwards, and scheduled git gc to avoid surprises at 2AM.
I mean, this isn't a 'URL returned error: 500' situation for anything that Codeberg provides considering this is an issue with Copilot and Actions.
Except actually it was, that was what my git client was reporting trying to run a pull.
I'm going to trust the constant stream of updates from the company itself which shows exactly what went down and came back up rather than a random anecdote.
I only found this post because I decided to check HN after getting HTTP 500 errors pulling some repos.
If you look at the incident details it also claims most services were impacted.
> Git Operations is experiencing degraded availability. We are continuing to investigate.
https://www.githubstatus.com/incidents/n07yy1bk6kc4
Recent years have shown this to be the wrong prediction strategy. The reason seems to be an incentive imbalance where there are quite a few reasons for companies to lie (including their own CLAs) and not a lot of repercussions for doing so (everybody competes on lock-in, not on product). Of course, the word-of-mouth approach is also exploitable by dishonest actors, but thus far there doesn’t look to be a lot of exploitation going on, likely because there’s little reason to bother (once again, lock-in is king).
This seems intelligent, after all companies are incapable of making errors in reporting and also have absolutely no incentive to lie about stuff like that. Those 500 errors others have reported as experiencing must have just been the wind.
I used to use codeberg 2 years ago. I may have been ahead of my time.
I rarely successfully get Codeberg URLs to load. Which is sad because I actually would very much like to recommend it but I find it unreliable as a source.
That being said, GitHub is Microsoft now, known for that Microsoft 360 uptime.
I have never had this issue. IIRC Codeberg has a matrix community, they are a non-profit and they would absolutely love to hear your feedback of them. I hope that you can find their matrix community and join it and talk with them
Actually here you go, I have pasted the matrix link to their community, hope it helps https://matrix.to/#/#codeberg-space:matrix.org
> Microsoft 360 uptime
I mean... It's right in the name! It's up for 360 days a year.
I mean... you understand the scale difference right?