One of the things that people should know about SourceHut is that even though it is "Free Software", it is not a very "free" platform. Certain types of projects are a red line to the owner and will be removed[0].<p>I'll be honest, I do not trust cryptocurrency stuff and don't believe that it has utility. I hold none, I don't use it. That being said, this makes me wonder what the next category of projects that SourceHut will ban is.<p>What is the point of a free software forge if you can't host certain types of (legal) projects? It's hard to take a platform seriously, to move all of your work there when the administrators can (and most importantly, actually do) ban your projects based solely on their personal views.<p>[0]: <a href="https://sourcehut.org/blog/2022-10-31-tos-update-cryptocurrency/" rel="nofollow noreferrer">https://sourcehut.org/blog/2022-10-31-tos-update-cryptocurre...</a>
Having tried GitHub, GitLab, Source Hut, Codeberg, and BitBucket (RIP), I think the only serious options are GitHub, GitLab self hosted, and Source Hut.<p>GitHub owned by MSFT aside, and also putting their AI push _also_ aside, GitHub has things like Sponsors, Dependabot, and code scanning that you won't find anywhere else. They sure are luring you into ecosystems that require a huge escape velocity, but GitHub serves as less-friction entry point to open source software.<p>GitLab self hosting is also nice and flexible, but it has a more "enterprise open source" feeling to it. Debian and Drupal can take advantage of it, but I don't see myself hosting mini projects like my mini CSS library that is also used by three other strangers.<p>Source Hut is really nice too. Admittedly I don't like email based workflows, so I haven't really thought about moving anything there. But Source Hut has "unlisted" repos, which I thought was really clever and missing from other forges.
I'm surprised the author rejected using GitLab because of its Open Core model, but decided foe using Drone CI, which appears to be following a similar model.
I was hoping for an evaluation of Radicle--the only forge that seemed to be doing something exciting and "new", and one I haven't had the time to really analyze yet--but... no such luck :(.
I also checked out SourceHut, but they're unfortunately not compatible with my email provider, and after some discussion with Drew (the founder) it started to sound like I'm not welcome unless I bring a real legal name and human identity. I understand how that type of collaboration can be valued and desired, but it excludes me, so I didn't pursue SourceHut any further. Something to think about if you ever consider it - no pseudonyms, no custom emails, use your real name and personal email address, or you'll be actively rejected by their software.
> Rather than needing a complicated YAML file to run a CI system, it’s just cloning Git repos and running commands.<p>How do you define what repos you need and which commands are run? Ah, it is indeed <i>also</i> a YAML file: <a href="https://srht.site/automating-deployments" rel="nofollow noreferrer">https://srht.site/automating-deployments</a>
<a href="https://ayllu-forge.org/" rel="nofollow noreferrer">https://ayllu-forge.org/</a><p>> Hyper Performant & Hackable Code Forge Built on Open Standards. Ayllu is a lightweight code forge designed to enable individuals and community projects to develop software in collaboration across open internet standards.
If your goal is to put some code out there just to share it with others, any of those will work. There's also Bitbucket, Sourceforge, and self-hosted Fossil. The blog author seems to be looking for more of an "enterprise" open source option. You could even use Gitolite or GitWeb if you're willing to self-host.
I’ve been moving into a self-hosted direction for a lot of my stuff as a lot of it really wasn’t being used by anyone but myself anyhow. My latest two projects are darcs on my home server using just SSH+Nginx for HTTP with an apply hook that run `nix flake check` & on the server end, a hook that runs `nix build` which not only checks the build, but also has the derivation in Nix store for me to pick up later as that server doubles as a substitutor. Collaboration can still be done via email of patches or DMing a pull request from someone else’s remote repo. It’s crazy how much we’ve been defaulting to tools that are so complicated--especially ones that require accounts to participate & have a social media features tacked on everywhere to cause the sorts of social media anxiety it does everywhere else. When the amount of collaborators is low, self-hosting is hardly a barrier that needs extra scaling.<p>Weird the author said email was a pain, but last patch I received it was as easy as `himalaya read --raw X | git am` to apply & pointing out that modern email clients do a bad job supporting this seems more damning of those clients than the workflow.
This post bothered me.<p>Perhaps I'm being too nitpicky, and there's nothing inherently wrong with their final choices, but their criticisms of the other systems alongside their reasons for choosing what they did are contradictory; overall the article was internally inconsistent.<p>Gitlab was dismissed for the SaaS being only open-core, despite opting for self-hosting in the end. Codeberg was dismissed for having limited CI, but in the end they went for a separate standalone CI anyway (& almost the same platform as Codeberg, just with the added overhead of self-hosting).
Coincidentally, I've been mulling the same idea for a week now. A self-hostable source forge + registry to serve artifacts would be amazing.<p>I think the only thing you'd lose is the 'social' aspects of GitHub, but I believe that can be made up with a combination of rss feeds (?), ActivityPub etc.
The author is too quick to dismiss Gitlab IMHO. I use Gitlab on the job, and Github for my hobby stuff, and Gitlab feels in many ways like a Github for professional use. The Gitlab CI system alone is a difference like night and day compared to GH Actions.
> Let me be clear, GitHub is still far and away the best website for open source discover<p>We need to separate discovery from hosting across the web. It's convenient to implement them together, but certainly not necessary. Bring back the link aggregators!
This is one of the first times I've seen someone who has casual access to Kubernetes.<p>They spent a lot of the post saying how they didn't really want to run stuff,<p>> <i>However, I’m a software developer, not a sysadmin. I want to spend my time developing software, not putting out fires and paying AWS bills for the rest of time.</i><p>But then their friend convinced them to "kubectl apply" a couple of manifests & they got a bunch of stuff spin up quickly, on what should be a reliable-ish maintained-by-other-people cluster. This is the first time I've seen a win like this & it feels like such a sweet spot!
This seems pretty filled with arbitrary back-reasoning for decisions they've already made based on guy feelings.<p>They don't like "AI-powered" because it is a vague, ill-defined term... except it's extremely obvious and well defined what GitHub means by that (and they've even used copilot and liked it!)<p>Gitlab CI is rejected because it uses YAML for config... like every other CI system they mention.<p>Gitlab isn't ok because it's open core, but Drone CI is.<p>Could have been an interesting comparison but it wasn't.
I’ve kind of anticipated this sentiment and have been thinking about providing Gitea hosting. Though I’ve been a little discouraged by Gitea’s release of their own cloud hosting service.<p>This blog post and the revenue growth at Gitlab over the last two quarters appear to be evidence of people wanting to moving away from GitHub (though their earnings report attribute the growth to other initiatives, clearly I had no real insight on this). As an extension of my self hosting of Mastodon efforts from last year I rolled up my other minor hosting efforts into <a href="https://hawt.cloud" rel="nofollow noreferrer">https://hawt.cloud</a> as a place to experiment with providing services like this.<p>This all ties in with the “Open Web” or “Small Web” sentiments I’ve seen expressed recently. The hyperscalers are locking more and more up into their ecosystems but there should still exist small vendors/hosting/service providers. They’ll never scale or be as reliable as the big guys but if they cease to exist a lot will be lost.
I just went back to kind of "anonymous ftp" via my gemini site on sdf.org. I find it far easier to maintain. When I make a change, a script packages up the change and sends it to sdf.<p>My github repositories were changed to point to it. The latest change by Microsoft made me move, I was considering a move when Copilot became a thing. 2FA got me off my butt to finally move.
One time I wanted to test some specific flows with Gitlab, so I had the stupid idea to try to run my own self hosted instance on a Digital Ocean pod.<p>It was not really complicated to install but the resources that it is using are insane.
First, it would not even start if you don't have a dozen of Gb of ram available.
Also, even if completely empty, the thing would take 30 mins to warm up doing I don't know what before being available for initial setup.<p>In honest opinion, trying so much to copy GitHub, Gitlab became a heavy bloated software stack.
I really wish a cross site standard emerged to enable PRs from one forge to the next without creating accounts:<p>- clone repo from original forge, push on yours<p>- open PR on your forge, kicking in a like-blog-trackbacks-of-yore API so that the PR would be known/opened on the original forge.<p>The thing would really put the "Pull Request" back in PR.<p>Lots of things to define and handle correctly to prevent abuse but it could be a much better distributed experience.<p>I guess big players like GitHub would nit like that too much though as it would fragment their coding social network.
Given the rise of “AI-powered” forges, i.e. platforms that add the source of projects hosted there to their AI models, is it ok to upload GPL-licensed code there?
> I hope they haven’t started down the slow and painful process of enshittification by following vague, ill-defined industry trends!<p>Has anyone else found themselves wondering if GitHub switched to a different web framework?<p>I feel like they moved further down the client-side rendering rabbit hole recently. Their UI feels increasingly janky to me. Several times a day I get a view of completely un-styled HTML that eventually organizes itself into the appropriate presentation.<p>For many scenarios, the GH web UI used to be preferable over navigating through visual studio because of how responsive it was. Now, I find the exact opposite to be true. I avoid the GH web UI as much as possible. VS2022 isn't great, but at least it's not implemented using react on top of some bullshit chrome wrapper.
> the email-based workflow was a lot clunkier than I expected<p>> If you want to create a repo for PR’s sake, email me at dev at notgull dot net and I can set you up.<p>So is there no way to allow seamless forking like on GitHub without high risks of abuse?
He doesn’t want to use GitHub because it trains Copilot on f/oss code, but then selfhosts and mirrors all his code to GitHub anyway?<p>Why release free software if you don’t want others to use it?<p>Why provide any code to GitHub if you don’t want GitHub to use it?<p>This is just tribalism, the ideology isn’t even consistent.
It’s hard to build a community around an open source if you’re not on GitHub imo. And if you’re not building an open source, than iiuc GitHub can’t use your code for copilot training so the main reason for not choosing GitHub according to the post isn’t relevant.
I think it's time we collectively mature past the point where taking unsubstantiated pot-shots at tools we don't "love" is seen as a value add to commentary.<p>In this case calling k8s a "nightmare" is a strong red flag akin to the mantra of "java is slow". It lacks the nuance of hands on XP and reeks of rules of thumb miss applied to the point of catchphrases.
Is not hard for me to see a case being built around the idea of finding value in having code to stay excluded from the surface exposed to AI so it becomes impossible for it to predate it.<p>It's a type of "security" that doesn't find an answer in "better AIs" or "AIs with improved alignment". The value comes out of being safe from the risk that can come out of the humans in control of these AIs.<p>I comes out of staying away and making impossible for any and all AIs to touch it.<p>Like the difference between food from a chef and hyperindustrialized food.<p>What if you want chef food only?<p>In the same way, exclusively human software creations.