IPv6 binds on wildcard addresses include binding to the IPv4 addresses.

What does ss -tlnp return? Does the process listen on any ports?

That should only affect ports below 1024.

Your two bind addresses might be in conflict with each other since [::]:5234 includes binding to the first one.

Python packaging and stability is a total mess. It has gotten to the point where I just look for alternative tools when I find out something new I found is written in Python.

You want https://tabby.tabbyml.com/ instead of tabby.ml

the info required was there already, just you needed to put effort in

Not really. This is mostly what this is all about. The companies are insisting that open source projects should do analysis of security impacts in addition to fixing the bugs whenever some “security researcher” runs some low effort fuzzing or static analysis thing that produces large numbers of bug reports and assigns CVEs to them without the consent of the project. The problem is that such an impact analysis is significant effort (often orders of magnitude more than the fix itself) by people with deep knowledge about the code bases and only really useful to the customers of those companies who want to selectively update instead of just applying all the latest fixes.

Talking about PRs being broken and then bringing up email, just about the most broken technology still in wide-spread use, is sort of ironic.

While true essentially forking the latest stable version of the kernel to make an LTS branch or a vendor version only multiplies the problem, it also does not contribute to solving it.

Yeah, the whole commenting won’t work if the server where the repo is hosted fails or the server where the person has an account. There is no redundancy.

I could be up and running in like 10 minutes to install Forgejo or Gitea

You could maybe do that but only because you already know how unlike most developers and you completely dismiss any active maintenance like updates, moderation, debugging performance issues, resizing storage,…

The term “single point of failure” means that only that point has to fail for the entire system to become unusable. You can easily have more than one of those in a system though.

Forgefed seems to be ActivityPub based which, judging by Lemmy, doesn’t solve the redundancy issue at all, it just allows you to interact with the content hosted in a single place from your own single place, giving you two single points of failure and two points where you can be tracked instead of one. This is not really the same kind of distributed as git repositories.

Can you name an open platform that actually does distribute PRs and issues? I know there were a few that tried but I mean one that actually succeeded and is usable by people who just want to report a bug?

Also, your issues and pull requests are much more likely to be lost in your self-hosted one project instance than on GitHub if anything happens to you.

I can understand the argument against GitHub in two contexts, one is when people build features into their software that assume GitHub, e.g. when a programming language assumes it can just prepend github.com/ to your repo to find it and the other is the argument that losing GitHub would be a huge blow because so many projects are there and only there so a lot of things would have to be done at once if that ever happened.

Federation doesn’t really solve the issue that self-hosting takes effort away from working on the actual project.

Haven’t used it myself but you could give https://rustdesk.com/ a try.

The worst thing is when it happens in this way and you can’t remember even though it was your own question https://xkcd.com/979/

Yeah, but when was the last time you decided to upload hardware device data for a root server to some hardware survey? That is something almost exclusively done by the kind of people who want to show off their system in some way.

I wonder how representative that is of actual software used. I would imagine hardware probes are run from installers and live systems quite frequently. I would certainly not expect several percentage points of “neither” in practical settings.