I’ve been looking at Codeberg but I’m really anxious to leave GitHub Actions behind and Codeberg’s replacement doesn’t seem ready yet.
We have a more advanced PR flow (stacking, round-based reviews), jujutsu support and we just launched our new CI system. Come join! https://tangled.sh/signup. The goal is to be the new town square for collaborating with friends and open source communities.
It's built fully in the open (https://tangled.sh/@tangled.sh/core) and we have a neat little community built around it on our Discord https://chat.tangled.sh.
Last night was a ton of fun with the sudden influx of users thanks to the GitHub news. :)
What does that mean? Doesn't jj just use git as its backend?
I've never used Gerrit directly, but I'm guessing that Jujutsu's change IDs are like a native version of what Gerrit does there.
I regularly (attempt to) do stacked diffs on GitHub with jj, and even with a bunch of automations it's still a clunky, awkward experience.
$ git cat-file commit caa00e20
tree e8418bf7e7c626fdf5b1570e6efcddf1e1b0c923
parent 1e3a841997f81c9b0ed21dc5a91e783fcf2f4ec7
author Anirudh Oppiliappan <anirudh@tangled.sh> 1754996643 +0300
committer Anirudh Oppiliappan <anirudh@tangled.sh> 1754996799 +0300
change-id skwlyzyomnwqntksmoqkrllqlpsskztt
xrpc/{errors,serviceauth}: split shared xrpc code to top-level package
Signed-off-by: Anirudh Oppiliappan <anirudh@tangled.sh>A few things that would make me actually use it: - namespaces (aka, organizations in GitHub terminology) - private repos (I usually start with private repos and then make them public)
The interface is really friendly and I like how approachable it is. You should make dedicated docs site instead of throwing folks at a bunch of markdowns. Maybe create a “Pages” equivalent and demo it by building a docs site?
Appreciate the feedback! A dedicated docs site is in the works (along side a pages offering!).
I'm not familiar with ATProto; what does this mean in practice? Does it just change the failure modes if something breaks?
You can have 1:1 parity with any company or product, but unless you have their word of mouth distribution and adoption, you will lose. Every time.
People in aggregate, as crowds, are relatively static and inflexible. Once they learn a fact once, it sticks. You cannot unteach that without lifting metaphorical mountains. The first mover with escape velocity wins.
The amount of energy needed to undo that is massive.
You'll have fringe 0.01%ers adopt some other tool, but they'll never carry enough gravitas to bring the entire network with them.
Anecdotal evidence:
- Github, Facebook, and Reddit have never been unseated
- Instagram has never been replaced, only supplemented
- Twitter/X has only lost steam due to extremely bad press, an unwanted name change, and a huge effort from Meta (which leveraged traffic and synergy from Instagram). And even then, it's still well within the public zeitgeist. Bluesky and Mastodon didn't even make dents.
- Google has never been displaced (granted, Google pays a lot of money to maintain defaults and maintain a web "pane of glass" monopoly, redefine the address bar as a search bar, etc.)
Just because I use Codeberg doesn't mean Github should die. Just because I never ever visit Facebook doesn't mean you should stop using it.
Why can't we all just play nice?
This is why VCs won't fund consumer. It's frothy, fickle, and the incumbents have saturated the network effects market. And that totally sucks.
I'm a believer in strong antitrust and that a fire every decade or so clears the forest of dead growth and renews the ecosystem. Evolution and reinvention staves off ossification. Instead of working on hyper-optimization of things like better ad targeting in search of more growth, we're working on more important problems and lifting bigger weights.
More startups means better compensation for ICs and innovators. Rewards for capital risk rather than extremely large wealth funds and pension plans.
Look at your Fords and Boeings. Totally stagnant. Newer international competition is refreshing and energized.
Or Google of just two years ago, before they had their "come to Jesus" moment. Where they witnessed visions of the loss of their single largest cash flow flashing before their eyes.
Except GitHub, maybe.
Granted, the replaced services were nowhere near the popularity levels of the current ones, and have made horrible decision.
This was pre-broadband penetration, pre-smartphone. That was a really different era.
Social networking wasn't really that big before Facebook. Friendster, MySpace, and Xanga were all niche communities.
> Except GitHub, maybe.
Github toppled SourceForge, which was awful.
Reddit is probably the player that unseated the closest peer competitor. It toppled Digg, which was pretty big at the time. (Digg had itself disrupted Slashdot, StumbleUpon, and Del.icio.us, but those were niche communities.)
https://tangled.sh is the other contender in this space that I know of. uses atproto (same as bluesky) under the hood.
fundamentally activitypub is insufficient to define these kind of networks. you'd need to have some sort of object-capability representation. the creators of forgefed are also moving in this direction: https://codeberg.org/Playwright/playwright
At some point I made very tiny contributions to OSS projects that had their own Gitlab hosted instances.
Sure a password manager makes it tolerable, but what about having an anonymous way of opening up PRs (subject to owner moderation of course)?
Use the author name and email for a virtual identity and when the PR request is accepted (not merged) force an email address validation for the PRer so that comment interaction can happen via email.
Fork the original repo onto your forge. Your forge could be your own forgejo instance, your own GitLab instance, just a plain git repo, or even Github if you like it so much.
Then send a note to the original authors informing them of your fork and the patch you wrote. Request that they review your work and pull it into their repo if they approve of it.
You don't need an account on their forge and they don't need an account on your forge.
Disregarding the fact that I would need to have a way to reliably find the contact information, the review process would probably take place outside the source code platform.
You have a point about the review process. How much of a moat do you want around your project? You want to keep spammers out. But that creates hurdles.
I was trying to interact with the Trunk project the other day. But they required me to create accounts on multiple services just to get in the game. When you need a Github account and a Discord account and... You know what? Nevermind. It's too much effort just to be nice and inform you of an issue with your project.
I totally understand that putting a public email address out there creates a huge burden for them. I suppose if I value a product enough I will accept some smaller burden and take the load off them. But most of the time I don't value a product enough to allow the authors to offload their burden onto me. Conversely, they don't value my input enough to accept the huge burden it would take them to allow me to interact with them.
The forge could automatically figure out if the submitted repo was a fork, figure out the sequence of commits in the foreign repo that diverged from itself, analyze the commit comments and commit diffs to see if they jived, and add the PR if it passed the spam filter.
Then you could submit PRs without an account or even having the contact information of a maintainer.
My top-of-funnel is not searching github but recommendations or searching technology/platform specific repositories e.g. for software it's flathub/f-droid and for rust its crates.io/libs.rs.
Where the code is hosted is in theory irrelevant... but I'm ashamed to say that when code turns out to be on gitlab my heart sinks. It's a bit of a red flag for e.g. no bug-tracking, no contributions, no maintenance, absent maintainer and unexpected licenses.
It's gross personal hypocrisy because I hate the absurdity of commercially owned FOSS collaboration and centralised git and happily self-host myself... but those not publishing code on github are awkward bastards :)
Similar for indexes of open source repos - I dont know of a singlar winner now but it was for a bit freshmeat for a feed of open source project updates
Public projects: Github. Having no problem with it. Also makes me happy poisoning LLMs with my shitty code.
I wanted to use their pages service as well to serve an SPA but their https://srht.site/limitations prevent SPAs from contacting external services I need. I get why they do that but I need my SPA to let users login to their databases and there's simply no way to do that while adhering to SourceHut's policy.
Fortunately pico.sh², codeberg³ and GitLab⁴ (not GitHub) don't have that restriction. I experimented with each of them last year. All of them worked reasonably well. Eventually I settled on GitLab which had the nicest CI/CD of the three at the time.
It’s not a direct equivalent to pull requests because it lacks all of the easy history, inline commenting, comment status tracking, and ease of use, among other things.
- GitHub for things I expect broad collaboration on
- SourceHut for things I just want to share (or expect contributions from a specific group that are comfortable with email git flows)
- Self-hosted for everything else. This used to be Gitea, but I've recently switched to charmbracelet/soft-serve, which fits my needs well (it's small and comparatively simple)
As other folks have noted, the social features of GitHub are hard to replicate elsewhere, but I've enjoyed SourceHut's stripped back approach.
Gogs is FOSS but basically BDFL. What does and does not make it into Gogs (it's still around) is ultimately decided by one person, and he's fairly conservative. Gogs is very fast but lacks a lot of features that would allow it to go head-to-head with GitHub or Gitlab.
So, Gitea was forked from Gogs to allow it to take a different direction, with a larger group of maintainers and more input from the user community.
There were two major attempts to have hosted Gitea. The first was Codeberg, a nonprofit based in the EU. The second is a business that took the name Gitea, is based in the US, and changed Gitea (the software) to the "open core, closed premium" model.
This change led to the creation of the Forgejo fork, which Codeberg adopted.
Hopefully you'll forgive me for shilling my own project. It's a source code sharing site with a few primary goals. First, don't break the back button. I started the project right around the time github broke too many critical features in short succession. The back button, the URL bar, and ctrl+f in the code view. I also want it to be easy to use as a federated collaboration tool. Ideally you'd start your own instance locally, get a familiar GitHub like interface for submitting patches anywhere, even by email. Or if you're hosting a project, you could have an always up instance that others could connect to.
The part I've been thinking about deeply the past few days is how to improve the discoverability of peer repos and forks. I want to create something github like in terms of collaboration, but also try to incorporate some of the best lessons from mailing list based repos where it's easy to grab and try patchsets (exposed as branches) from a 'fork' without losing the value of a cannon 'upstream' and without insisting that any specific upstream is where every single commit belongs.
The whole thing is written in zig without any dependencies other than git for some of the repo management features I haven't ported yet. (and for generating git blames) If you do use it, or notice any issues, or think it's missing features do let me know. I'm currently trying to decide which is the next most important thing to hack on :)
see also https://github.com/GrayHatter/srctree if you want to subscribe to updates (still a feature srctree lacks lol)
More:
- Annual financial reports: https://disroot.org/annual_reports
- Someone's review: https://proprivacy.com/email/review/disrootIf you're using GitHub for your public projects, why not just use GitHub period? Genuine question here.
Slow, simple, inexpensive, safe, and good enough.
I've always been fascinated by it, but it seems to be completely out of the limelight.
Committed items are stored in 2 places:
1. anon ftp on sdf, back to using tar files via gopher, the main site
2. gitlab is a mirror and will make it easier for youngsters. FWIW, I find gitlab easier than github.
If gitlab starts going the way github did, I will delete my items on gitlab like I did on github a long time ago.
OSS work is mirrored to Codeberg and SourceHut. For actions I try to make sure that local builds, and cross compilation to Windows, macOS, armhf and arm64, is always working to not soley depend on Github Actions.
So, given that this will happen everywhere, does it really make a difference what MS does?
Still browsing. https://radicle.xyz looking rad, though.
I've moved the bulk of my repos from GitHub to cgit first, but now to https://tangled.sh.
Where should I host software for individual papers in 2025 now that GitHub is part of Microsoft AI? @ academia.sx
https://academia.stackexchange.com/questions/220795/where-sh...