Microsoft shot itself in the foot by not promoting Azure DevOps more, and avoiding the wrong "Microsoft-only tech" impression. The product is far more mature and simply works as expected (for the most part), while having everything in one place. The only area where it's lagging behind now is AI integration. Given that it's still a very much viable alternative after years of under-investment in favour of GitHub is quite telling.
Has a solid CI system that is quite easy to pick up and comes with a bunch of nicely integrated features, such as Container and Package registry, Terraform/Tofu state management, K8S cluster integration, and more.
Moving from GitLab CI pipelines at my old job to GitHub pipelines at my new job felt like stepping back in time to the Stone Age. So much stuff in GitHub overall that just totally sucks that I don’t understand because it must be one of the most dog-fooded services on the planet.
Sure. Most egregious to me because it’s such a simple usability thing (I was able to fix it myself with some custom css): when viewing a list of PRs, the approval or changes requested status is a tiny little grey text-only label that blends in with all the other grey text. Makes it very hard to see at a glance which PRs are approved vs changes requested vs awaiting review.
Next is being unable to configure a manual PR pipeline job. In GitLab it’s as simple as when: manual (I think, it’s been a while) to configure a pipeline that is associated with a PR, but requires triggering manually. I might want to do this with e2e or mutation tests for example. I want them to still run & require passing before the PR can be merged, but I don’t need them to run on every commit, just once at the end before merging. In GitHub I don’t think this is possible, pretty sure workflow_trigger doesn’t associate it with the PR. I’ve managed to come up with a hack that detects if the pipeline job is a manual re-run and that will have to do haha.
Lastly, GitLab has much better (or actually exists at all) automated test integration. It comes with a built in test results browser, and built in test coverage tracking that can automatically track the change in coverage between the PR and main & show that on the PR, block it if it decreases, etc. Even can show the test coverage in the PR diff!
It comes from the phrase ‘eat your own dog food’, which basically means being able to test your own products by actually using them yourself. Here is a link that can explain it better than I can https://en.wikipedia.org/wiki/Eating_your_own_dog_food
Surely every developer at GitHub uses GitHub themselves for their work, so they must experience all the annoying little things, and yet those annoying things still exist
I like to think the obtuse name is some kind of warding against people with hopes of making money off it and bastardizing the project. The name Forgejo is functional in that it is unsellable.
In the current internet environment, I don't imagine many people read a vague "political reasons" in the broader sense of organisational power dynamics.
"Political" does not automatically mean "bad" or "invalid." It was a while ago, and the engineering effort is there. Simply using your own tool to develop the tool goes a long way.
Ironic that a low-effort, one-word, drive-by comment is now upvoted, while actual discussion is not. As if simply saying "forgejo" around Gitea discussions is supposed to mean something.
Anyways, dogfooding and having LTS releases made Forgejo preferable to me. Moreover, we have agents now. One can literally ask to clone both and compare commits for the last year on subject and size to get a better idea of where things are going and how fast.
It really is a beauty. My employer used to use an ancient version of Gogs until I came along and stuck Gitea in their faces. Now we use it for everything. Issue tracking, public and internal. CI. Wikis. Debian repo where we were previously just building deb packages and manually rsycing them around + dpkg installing them.
gitea's development is hosted in github and there doesn't seem to be any gitea mirrors of it. forgejo is basically gitea but better and it's actually developed using forgejo.
People keep telling me this but I've looked at their comparisons in their docs and as far as actual technical differences, not just vague political arguments, I see nothing compelling enough to convince me to use it.
It's also insanely bloated using multiple GBs of memory for a fresh instance straight out of the box.
Eh, that's not really something a company would be bothered by. Small instances (up to 1000 users) can run on a 8vCPU/16GB memory VM which isn't much of a dealbreaker.
I remember some drama about them rejecting feature PR’s for the free CE that overlapped things they wanted to keep locked behind the paid EE. This was a pretty long time ago, but is that not still a concern?
On a practical level I’m not saying it’s a bad option. I’d even agree it is one of the best free options for many cases as you do get a lot out of the box as is. The stewardship of a piece of oss is always important to consider especially when it’s a full ecosystem commitment that can become a complicated migration to leave.
I’d just say if you are going to invest time heavily into the ecosystem, be aware you might eventually have to go without something, budget a “speak with sales rep” amount per seat, or migrate your operations and project management away from it.
Yea seems like they are all about the openness of everything which I understand but at the same time it doesn't really look like they are trying to directly compete with github in that aspect.
Codeberg. I've also looked a bit at tangled.org, where you self host your own node, but it gets tied into a single network, so it all looks like one app
Recently moved to self hosted Forgejo. It's fucking slick. Still waiting on federation support but for my own shit, it's great. I still am on Github because it's basically social media for developers. But for my own projects, I host them locally there
Despite what they might think, GitHub cant be the hub for agentic coding workflows if they cant get the basics of being a git server right. People will leave if the product is trash and the alternatives are just as good
My personal usage patterns of GitHub haven’t changed since fully migrating away from SVN nearly 15 years ago. Though I’ve been slowly migrating to my own gitea instance more recently.
That is smart. The git data does not matter as much to me as the files. I don't really do any retrospection of git commit logs beyond resolving the latest git tangles I get myself into.
Seriously, keep a github mirror. Historically, all personal websites and self-hosted things go down within a few years (usually because of a loss of interest/life events, and no, reader, you won't be different even though you're all hyped about self hosting right now).
Github will still be there, in one shape or another. Keep a read-only mirror of all your FOSS projects there. Write in bold that this is a mirror and try to convince them to contribute to your self-hosted instance instead (they won't).
I've held onto the same 2 or 3 domain names for 20+ years. I just need to find something low cost, isn't AWS, and isn't going to fold for at least the next decade.
I understand the platform risk, even the platforms I host and manage myself.
I personally have a daily borgbase backup that I download and check twice a year.
I'm not even hyped about self-hosting. It's just cheaper than 12 different cloud services that are potentially stealing all my IP because it looks like copyright law means nothing if it's for your AI empire.
I would agree... except right now what we're seeing is the death of github, if you can't even trust a merge, then it is worse than unusable. If this was the 70's, this post would be titled "Github considered harmful" and it would be 100% right.
There's only 1 website I trust to outlive me and everything, and that's sdf.org
It's fun to poke fun, but there's a world of difference between being a git server for a few codebases and being the preferred, free service for the entire world. Their infrastructure was not built for the amount of traffic they're getting hammered with, and we're all paying the price by tying our stuff up to their services. I guess in my case I don't pay, so I'm not mad, but if I did pay I would be.
It very much did not, and their massive user and org growth since the Microsoft acquisition, not to mention forced migration of various services from AWS and self hosting to Azure, were also contributors.
It’s also important not to have rose colored glasses here. GitHub has always been a home of many flaws in its different eras.
Believe me I don’t have rose tinted glasses, but I also remember never actually being annoyed with the state of things until this last year. This is the first time in my life that my literal job is being impeded by GitHub not working
I fondly-not-fondly recall 2016 being that for me. My team (.NET team at Microsoft, actually!) was all in on using GH for development and every day was a crapshoot of if a PR would load because we had the audacity to leave thorough reviews. We had them on the phone a lot and the team was responsive but our “lots of people leave lots of thorough reviews” workflow broke most of their architectural assumptions at the time. It took about a year to get reliable, but we persisted with it and supplement med with some home-grown tools as needed. I don’t think they’d ever dealt with such a density of activity surrounding the code at the scale we operated at (over 300 engineers), so the current times are just rhyming for me.
It defaults to 100% uptime, even if there is no data being fed in, so if they started tracking using that page in 2019 then everything before it would be by default 100%
https://www.githubstatus.com/uptime?page=3000
Including apparently 100% uptime in 1996! 10 years before the site was even created.
It's not about the amount of traffic, we just had to migrate from bitbucket to GitHub and it's atrocious how bad the user experience is, bitbucket gives you a nice overview of your PRs in progress and what to review
For GitHub one of our DevOps guys had to vibecode a greasemonkey plugin to do that, though you can get a chatgpt interface to ask it what you still need to review
They certainly are! Especially since things like /fleet in copilot CLI are literally designed to just swarm commits (each of which kicks off a CI run, etc).
My guess is they estimated they’d have a lot more runway to address things last year, as I have no doubt plenty of people internally knew this could happen. But they didn’t anticipate Claude Code taking off like a rocket last Winter. I’m sure a dozen or two SREs there are saying “i fucking told you so” in their heads every day.
world of difference between being a git server for a few codebases and being the preferred, free service for the entire world.
What ? Isn't coding solved based on microslop ceo statement . Why would they struggle with scaling,I am sure they have tons of free azure servers. Didn't microsoft layoffs 10000 employees just now because ai is 10x multiplier. why are they struggling
No they didn't. This is just nostalgia talking. GitHub was pretty much in a feature freeze state going nowhere when Microsoft bought it. I'd argue that if it hadn't been bought, GitHub would not be relevant today. No matter how much the doomsayers would love it to be otherwise, Microsoft saved GitHub and aside from a few well known fuckups, GitHub has consistently improved year on year under Microsoft's ownership.
Remember when you had to pay to have private repos? I do.
Yeah, but I think that was a perfectly reasonable line to draw. Microsoft didn’t make it free out of the goodness of their hearts, but for PR bragging rights, and now they have to make the money back elsewhere, in a more convoluted business model.
TBH I think Github simply threw away old status history when they migrated from status.github.com to githubstatus.com in December 2018 (conveniently shortly after Microsoft's acquisition closed).
Doing a Google search for historical Github issues led to an incident on March 2, 2018 which is listed with 100% uptime here.
That or Microsoft is being far more transparent about outages than Github ever was.
It's exactly this, although I wouldn't give MS credit for it. This graph actually shows the opposite of what it purports to. GitHub had tons of outages before the Microsoft acquisition, but didn't have the operational maturity to actually handle incidents and statusing in a consistent way. What appears to be more incidents post-2020 is actually an increased internal emphasis on incident communication.
Just jumping around the wayback machine, there are error messages present on other days too. Each incident has at least a symptom and resolution present. So while you may have a point about not communicating enough detail, the amount of downtime itself seems to have been clearly communicated on the old page.
While this looks damning, to be fair, there are many other things that happened during this tim3 that could at least be partially to blame for this trend.
Not defending Microsoft here, but this is an over-simplification at the very least. For example, most of the real activity in this graph happens starting in 2020 (covid times)
And also if you look at the breakdown you'll realise the vast majority of the downtime was GHA. Everything essential like core git operations and issues are still pretty solid.
Most of all was that after the Microsoft acquisition their growth really started to take off, and Microsoft pushed tons of enterprises to use GitHub over TFS and Azure DevOps. Just an endless stream of growth and scale across every dimension imaginable, now accelerated since everyone and their mother is letting Claude push code at scale.
This does not answer my question.
I see a lot of people that complain about services or their quality and they use free tiers and do not pay anything yet expect 24/7 support.
So he wasn't paying for a product yet expected premium features?. Giving your data away is part of not paying for something i guess.
I do not expect enterprise grade service for free solutions.
They are under no obligation of having an uptime of 99.9% for non paying customers.
You might like it or not, but it is what it is.
If you pay and get shit service, then you can complain and escalate, otherwise you just accept your fate.
I am not even asking for 99.9%, I am just asking to be able to use the fucking website. Nowadays when I try to use the site, it’s a 50/50 on if it even works. Not to mention that my employer DOES pay, and we still can’t even use the site half the time
Drew Devault, the man who enjoys sexualizing pre-teen girls, generously offering antiquated-by-design git hosting. Versus Microsoft, the company that will rape you and then lawyer-talk into agreeing that you had it coming, providing an increasingly bloated but good web interface.
GitHub usage is mostly about marketing and customer familiarity, people going to think twice about downloading software from a service they have never heard of.
Is it coincidence that GitHub's reliability has become progressively worse as agentic coding has become increasingly prolific? An exercise for the reader.
Yeah, I did. Read my other post here. I explicitly said what the article was about, and he's not leaving because GitHub is horrid. Did YOU read the article?
Idk man, constant outages that are blocking work sounds pretty horrid for critical infrastructure in CI/CD.
When you hit your nth day in a row of workflows being crippled because of bad service health, it becomes pretty hard to view things in a positive light.
They also expressed that they’re angry at the state of things and have been very vocal and mean about it. Calling it “horrid” in its current state wouldn’t be an inaccurate statement.
And no, I haven’t read your other posts here, there’s 200+ comments and I don’t always look at usernames. Get over yourself.
I think they don't do it anymore, but they upped subscription prices back in the day and limitted repos.
I already had my gitlab account so I moved to gitlab. Im also using codeberg.
Then there is the issue: github is closed source, microsoft owned. And MS cannot be trusted with FOSS. The fact that MS owns it, is a determent to the product in itself.
Maybe I'm out of the loop. What's wrong with Github exactly? I don't use it for git actions because it never appealed to me. But for code repository outside of I think 2 maybe 3 noticeable outages this year, it's been good to us.
So that was it, wasn't it .. i kept wondering if my laptop had slowed down drastically or was my memory failing me but GH pages used to load crisply. Now, the structure loads, some animations play out and then the data eventually renders. It's so irritating.
They've been having a particularly bad string of outages and general reliability problems since agenting coding really took off late last year. Far more than normal, and it's seeming like there isn't an end in sight right now, since "by design" behavior (like pull requests kicking off tons of work) are what are being stressed.
It’s a result of the Microsoft acquisition and then moving from AWS to Azure. When the foundation is cracked, everything that’s built on top of it is not stable.
They were for a while, but that was long resolved. The bigger issue was their own self hosted infrastructure and piles of code optimizing for that specific infrastructure.
I mean...did you read the article? The author pretty clearly outlines what's wrong with it.
That said, if you want empirical evidence, check out the unofficial outage metrics page. 87.5% availability for the platform, with most days having at least some type of outage.
Since Microsoft bought them, they have been unreliable. Their business decisions and practices are highly questionable and people want nothing to do with it
It's all about github actions really - I find it amusing that to most people github actions is a huge part of github, meanwhile we're over here happily using the baseline source control features!
tbh I can't imagine mixing CI with my git provider, simply because I want my git provider to do one thing and do it really well - git it?
A ton of outages with github actions, timing reliability for github actions (Zig developers mentioned this), some more outages not related to actions, like the ones you've thought of, the recent (a month ago I believe) problem where merge queues deleted work.
Also moved all of my personal projects away from Github, the way they decided to do AI training on people's code (you have to specifically opt-out) was too much of a dick move.
GitHub ToS have you give them a licence to your code for any purpose and have you waive your right to sue them for it.
The ToS always were like this, but people did not like it when GitHub starting using for AI training.
I would bet most folks responsible for creating it, are no longer under Microsoft paychecks.
This is what happens with most acquisitions.
Xamarin one, also went down quite bad I would say. The only thing left of it, is the infrastructure used to target iOS, Android and WebAssembly. Everything else was either replaced by modern .NET, or rewriten in incompatible way (Xamarin.Forms => MAUI).
this post is like, way to raw and emotional for what seems like something that is just such a bandwagon right now "oh no github has 2 nines availability lololol". it isn't causing me trouble personall so maybe i'm missing the point but it's like, why get so upset when you can't hit refresh there sometimes. are their release workflows failing? maybe that matters. otherwise, barely registers
Ctrl+f release (0 results) Ctrl+f review (1 result, yes, he couldn't review a PR for 2 hours) description of him writing an X on a physical calendar, like he is a prisoner in a virtual github prison, counting down the days to his inevitable death...yes there is one result for that as well
What a moron lol...Good luck to him...I think Ghostty days are numbered anyways. I could literally recreate their platform in a matter of a month. Maybe I will just to prove my point.
Have you used GitHub in earnest? Please be honest. I’ve used both extensively and anything Gitlab does GitHub does better. And there are many things that GitHub has which Gitlab does not.
Can we be honest and say that no one wants to go making a detailed case for any reddit comment and all of us have vague opinions based on past experiences that we are not going to specifically dig up?
Fair enough, but it's also way more convincing if you provide examples. One thing to vent, but if you're trying to change someone's mind, I don't think saying 'trust me bro' is enough!
Could you provide some examples? Genuinely curious. I've been using GitHub for nearly a decade now and have just recently started using Gitlab for some particular things and it's been a breath of fresh air.
It's hard to list them from memory. They're things you notice as you use it.
Some examples are aggressive pagination on the "changes" tab of an MR. Aggressive collapsing of "large files" on the changes tab. Want to ctrl+f a specific string to see if it exists in the changes? Sorry, you can't because that change is on page 2 or in a collapsed file. Ok then let me open it in the "Web View" so I can see the full MR changes and search there. Oops, Web View doesn't support search yet. Meaning, there is literally no way to ctrl+f a string in an MR without cloning the branch locally and searching locally.
Linking directly to a line of a file fails at least 50% of the time.
The worst thing in my opinion is that MRs will sometimes open to a seemingly random specific commit in the MR with absolutely no visual indication that it did other than noticing it in the URL. When this happens, you may not realize you're only reviewing 1 commit from the MR and not the entire MR.
The revert button on MRs exists but is needlessly convoluted. In GitHub, you click "Revert" and it opens a new PR with the exact opposite of the changes in the MR against the branch you merged it into. In Gitlab, it has an incredibly convulted poorly explained flow that I usually just fumble through randomly when the reality is that I want it to do what GitHub does 100% of the time.
GitHub also includes the merge commit when you select "Squash & Merge". Gitlab does not. So the main branch has completely polluted history of pointless merge commits.
I think these are the main things. But they're issues with the core, basic functionality that I experience every day. Not weird edge case issues.
I use Gitlab for school and it’s been pretty decent so far. I use my GitHub account for personal stuff and I’m way more familiar with it but I’m ready to jump ship as well. It just sucks that private equity or big tech companies buy up all of these genuinely good and useful services and turn them to shit.
Like, we can go to gitilab or whatever other alternative but if it gets popular enough it’s just going to be the same thing all over again.
For fun, I have been toying with the idea of seeing what it would be like to use a minimal self-hosted setup. You can sort of get ci by simply using a post-receive hook, then you just display the logs. Once you have that, you can manage VMs by pushing Ansible playbooks. Now you have pretty much bootstrapped a infra-as-code setup and can have it host whatever you like, like your app or whatever you need for development (bug tracker etc.)
I think now, the viability of such a setup is real, compared to something like gittea.
No one so far has assumed this is because of the recent issue where the PR being merged ended up being put on a different commit, thus git history being erased and potentially quite a bit of time to untangle the mess. Having the UI telling you one thing and then merging with a random prior commit is a bad look.
The merge problem probably didn't help, but the article is pretty focused on downtime. It also mentions in the footnotes that they've been considering and planning to move for months.
Recent outages and the constant CLI/API hiccups make it hard to trust them with anything beyond vanilla git. How's agentic coding supposed to work on a shaky foundation?
I've felt this way for a long time, but for the past month I've kept a journal where I put an "X" next to every date where a GitHub outage has negatively impacted my ability to work2. Almost every day has an X. On the day I am writing this post, I've been unable to do any PR review for ~2 hours because there is a GitHub Actions outage3. This is no longer a place for serious work if it just blocks you out for hours per day, every day.
I get why people are annoyed with GitHub, especially after the outages. But GitHub is still where everyone is. For FOSS, it’s still the default place where things happen.
AI slop is real, but moving won’t magically fix it. If another forge gets popular, the same garbage will show up there too.
Lock down PRs, restrict comments, require approval, whatever. That seems less painful than moving everyone elsewhere. Just please don’t pick GitLab :( It’s slow enough that I often give up before contributing.
well, I guess when u put it like that, nothing will ever change and everyone will always stay on gh forever. in fact, im posting this comment on slashdot at this very moment!!
Who cares? Throwing out a few hours of lost productivity here and there to just dedicate dozens and dozens of hours and money running your own solution. This is like the bootcamp CTO guy getting all that press moving away from the cloud a few years ago.
Apparently YOU care? Also, nowhere on the article did he mention doing it "in house"... and as a public open source project, I doubt he is buying anything anyway. Why are you so outraged that a project wants to move to a new hosting service after frustration with the existing one?
Just track the engineering effort to migrate and refactor workflows and permissions. I hope he finds the effort to migrate and the new provider’s availability is an enabler over the GitHub outages.
Popping in to say the article was clear, concise, and honest. As I'm getting a bit older, I notice how much I enjoy honest writing with restrained brevity. It was a lovely read, well done (and frankly, the issues at GitHub are maddening to anyone touching repos hourly or daily)!
The amount of bullshit Github has put me through makes me very willing to use literally anything else, I'd prefer managing 100 logins for Git forges than use it atp
I mean, he talks about how much GitHub has meant to him, and he and Armon hired their CEO from GitHub, so yeah, I’m sure he would be back if it worked properly.
They need to first separate their infra from paying customers and free slip machines. Then have a consistent pattern of achieving a thing. It seems GitHub problem is multi layered. Actions, breaking UI, backend shots on large prs, massive artifacts and their whole wip ports to react
Even with the company I work at the outages have had a measurable impact on “developer productivity” isn’t something that’s explicitly measured.
We’ve even had to delay client onboardings a day here and there. Something we can’t afford to do considering that we’re a startup and any lost contracts would be devastating.
Any production repos I and my company have are not on GitHub. And that's perfect.
Only open source projects end up there. Usually for the exposure.
But AI has pretty much destroyed all GitHub usability.
Looking for a solution to a specific problem? Good luck with thousands of vibe coded projects with a single commit 3 month ago.
Do you have an open source project on GitHub? Have fun with bot generated PRs completely unaligned with the projects vision.
I can absolutely understand the motive and I wish the project the best of luck.
aventus13@reddit
Microsoft shot itself in the foot by not promoting Azure DevOps more, and avoiding the wrong "Microsoft-only tech" impression. The product is far more mature and simply works as expected (for the most part), while having everything in one place. The only area where it's lagging behind now is AI integration. Given that it's still a very much viable alternative after years of under-investment in favour of GitHub is quite telling.
mechaniTech16@reddit
I’m a huge ADO fan and you’re spot on. It’s mature and has so much to offer for the business folks too
scribe-kiddie@reddit
I've been using ADO but mostly for its Repos and Board at work and GitHub for personal project at home.
Personally I think GitHub is better in almost every aspect. What am I missing?
ilawon@reddit
First thing I noticed after changing companies and going from Azure devops to github was the atrocious code review flow.
Where are my comments?
OH, I need to submit a review for the comments to appear?
I see other issues but most likely related to how this company sets it up.
foramperandi@reddit
I assure you Azure DevOps would be down 100% of the time if it handled a fraction of the traffic GitHub does. You're comparing apples and oranges.
xill47@reddit
That assumes that ADO Pipelines and Github Actions run on different infrastructure
foramperandi@reddit
It'd be easy to prove if that was the case, since they'd have the same uptime and same outages.
xill47@reddit
Only if we also assume that they run the same unmodified infrastructure code (and Github has a few more features there).
NenAlienGeenKonijn@reddit
The argument being made here is that at least it's scaled properly for it's workload.
aventus13@reddit
I'm not saying it wouldn't but uptime is only one of the - increasingly complained about - problems that GH has.
Semick@reddit
Team was gutted and mostly moved to GitHub actions year before last. Its mostly a skeleton crew at this time. Actually sucks.
Windyvale@reddit
I’ve been deciding on an alternative myself. I think GitHub is no longer for developers.
Gabelschlecker@reddit
GitLab is nice (and quite common across Europe).
Has a solid CI system that is quite easy to pick up and comes with a bunch of nicely integrated features, such as Container and Package registry, Terraform/Tofu state management, K8S cluster integration, and more.
young_horhey@reddit
Moving from GitLab CI pipelines at my old job to GitHub pipelines at my new job felt like stepping back in time to the Stone Age. So much stuff in GitHub overall that just totally sucks that I don’t understand because it must be one of the most dog-fooded services on the planet.
silksong_when@reddit
Can you give any concrete examples?
young_horhey@reddit
Sure. Most egregious to me because it’s such a simple usability thing (I was able to fix it myself with some custom css): when viewing a list of PRs, the approval or changes requested status is a tiny little grey text-only label that blends in with all the other grey text. Makes it very hard to see at a glance which PRs are approved vs changes requested vs awaiting review.
Next is being unable to configure a manual PR pipeline job. In GitLab it’s as simple as
when: manual(I think, it’s been a while) to configure a pipeline that is associated with a PR, but requires triggering manually. I might want to do this with e2e or mutation tests for example. I want them to still run & require passing before the PR can be merged, but I don’t need them to run on every commit, just once at the end before merging. In GitHub I don’t think this is possible, pretty sureworkflow_triggerdoesn’t associate it with the PR. I’ve managed to come up with a hack that detects if the pipeline job is a manual re-run and that will have to do haha.Lastly, GitLab has much better (or actually exists at all) automated test integration. It comes with a built in test results browser, and built in test coverage tracking that can automatically track the change in coverage between the PR and main & show that on the PR, block it if it decreases, etc. Even can show the test coverage in the PR diff!
rzet@reddit
i used gitlab years ago and still cry each day they ask me too look at jenkins or github actions...
punkbert@reddit
What does that phrase mean? (English second language here)
young_horhey@reddit
It comes from the phrase ‘eat your own dog food’, which basically means being able to test your own products by actually using them yourself. Here is a link that can explain it better than I can https://en.wikipedia.org/wiki/Eating_your_own_dog_food
Surely every developer at GitHub uses GitHub themselves for their work, so they must experience all the annoying little things, and yet those annoying things still exist
SupersonicSpitfire@reddit
To be fair, they are both akward YAML.
ryanstephendavis@reddit
Agreed. GitHub sucks once one sees how easy it is to define CICD in GitLab
Leliana403@reddit
It's also insanely bloated using multiple GBs of memory for a fresh instance straight out of the box.
Gitea on the other hand is very small and has its own version of GitHub Actions so you don't even have to rewrite your workflows.
Ferilox@reddit
forgejo.
ferow2k@reddit
Ok. But couldn't they have chosen a name that was at least pronounceable?
Sitethief2@reddit
What are you on about? Forge + jo. The place a smith makes tools + the short form of the name Joan.
SirOldbridge@reddit
Double /dʒ/ is clunky to pronounce
jonpacker@reddit
If you think this is an intuitive name to pronounce you are seriously the first person I've ever encountered to believe so.
The first comment anyone has about Forgejo is how the hell you say it.
ferow2k@reddit
Right. It's so easy that they had to add phonetic and audio sample to the first question of their FAQs.
jonpacker@reddit
I like to think the obtuse name is some kind of warding against people with hopes of making money off it and bastardizing the project. The name Forgejo is functional in that it is unsellable.
Leliana403@reddit
You realise Forgejo is a fork and rename of Gitea, yeah? Think about that for a moment.
trannus_aran@reddit
For-JAY-hoe? I agree though
Leliana403@reddit
Which is just a Gitea fork made for political reasons.
chiniwini@reddit
Everything is political. The very existence of open source software (and thus github, gitea, etc) is political.
Hipolipolopigus@reddit
Politics is when the lead maintainer silently transfers the project, its trademarks, and its domains to a for-profit corpo.
Chisignal@reddit
I mean, it is politics. It just happens to be a really good reason for a fork
Hipolipolopigus@reddit
In the current internet environment, I don't imagine many people read a vague "political reasons" in the broader sense of organisational power dynamics.
SafePerformer@reddit
"Political" does not automatically mean "bad" or "invalid." It was a while ago, and the engineering effort is there. Simply using your own tool to develop the tool goes a long way.
Ironic that a low-effort, one-word, drive-by comment is now upvoted, while actual discussion is not. As if simply saying "forgejo" around Gitea discussions is supposed to mean something.
Anyways, dogfooding and having LTS releases made Forgejo preferable to me. Moreover, we have agents now. One can literally ask to clone both and compare commits for the last year on subject and size to get a better idea of where things are going and how fast.
Plank_With_A_Nail_In@reddit
Its 2026 no one cares about a few GB of memory anymore.
Leliana403@reddit
Have you not been paying attention to memory prices?
rusmo@reddit
I’ve recently started running Gitea on my home lab. I’m using actions but none of the issue tracking stuff yet. So far no complaints!
Leliana403@reddit
It really is a beauty. My employer used to use an ancient version of Gogs until I came along and stuck Gitea in their faces. Now we use it for everything. Issue tracking, public and internal. CI. Wikis. Debian repo where we were previously just building deb packages and manually rsycing them around + dpkg installing them.
You're welcome, now pay me more.
rusmo@reddit
Yeah, they always give ::surprised pikachu:: at this last part.
loveisnomorethandust@reddit
gitea's development is hosted in github and there doesn't seem to be any gitea mirrors of it. forgejo is basically gitea but better and it's actually developed using forgejo.
Leliana403@reddit
People keep telling me this but I've looked at their comparisons in their docs and as far as actual technical differences, not just vague political arguments, I see nothing compelling enough to convince me to use it.
Gabelschlecker@reddit
Eh, that's not really something a company would be bothered by. Small instances (up to 1000 users) can run on a 8vCPU/16GB memory VM which isn't much of a dealbreaker.
Iwisp360@reddit
Gitlab forbids access to Cuba
Status-Importance-54@reddit
Gitlab ist utter Trash. Mist of the features are build for mgmt slides
lolmycat@reddit
Gitlab’s biggest issue is how insanely expensive they make self hosting.
goldman60@reddit
Self hosting is free as long as you already have something to host it on
worldDev@reddit
I remember some drama about them rejecting feature PR’s for the free CE that overlapped things they wanted to keep locked behind the paid EE. This was a pretty long time ago, but is that not still a concern?
goldman60@reddit
Might be? I wouldn't personally contribute to a freemiun open source project like gitlab
worldDev@reddit
On a practical level I’m not saying it’s a bad option. I’d even agree it is one of the best free options for many cases as you do get a lot out of the box as is. The stewardship of a piece of oss is always important to consider especially when it’s a full ecosystem commitment that can become a complicated migration to leave.
I’d just say if you are going to invest time heavily into the ecosystem, be aware you might eventually have to go without something, budget a “speak with sales rep” amount per seat, or migrate your operations and project management away from it.
lolmycat@reddit
No it is not. Free tiers are not for real development teams.
goldman60@reddit
What does your "real development team" actually need from self managed premium or ultimate?
Only thing I use day to day is merge trains and that's only because there are 50+ people on my program.
pixel-der@reddit
I was also considering this, are there any good alternatives?
Houndie@reddit
I've been moving to codeberg. You'll have to get used to a huge reduction in features. Luckily, I don't need most of those features.
twigboy@reddit
Can you name some examples? Also considering for my private side projects
Atulin@reddit
No private repos, no discussions
Houndie@reddit
No suggestions on PR reviews. No app support. More difficult CI story. No web code editor.
WanderingInAVan@reddit
Codeberg
Crafty-Waltz-2029@reddit
Can I use codeberg and forgejo self host at the same time?
WanderingInAVan@reddit
Don't see why not. It's two different setups and honestly I prefer self-hosted solutions over centralized most of the time.
Its just duplicating work to make sure your code remains available. Not an unreasonable action to take in my view.
mok000@reddit
It’s a great name. Guess Cody McCodeface was taken.
ripter@reddit
https://codeberg.org/
zig and others have already moved there.
ray591@reddit
IIRC, Doesn't allow personal, private repos right?
helloworldpi@reddit
https://docs.codeberg.org/getting-started/faq/#how-about-private-repositories%3F
ray591@reddit
Yep, it doesn't.
helloworldpi@reddit
Yea seems like they are all about the openness of everything which I understand but at the same time it doesn't really look like they are trying to directly compete with github in that aspect.
LGXerxes@reddit
Perhaps after forgejo lands pub/sub codeberg can extend and offer private repo's etc.
TheGRS@reddit
GitHub was similar for a pretty long time. I think they only made private repos free after the MS acquisition.
Never_Guilty@reddit
Yup, I remember using gitlab because you had to pay for private repos
unapologeticjerk@reddit
This is correct.
hutxhy@reddit
Wait, what? I have a private repo on codeberg
th1bow@reddit
same
ray591@reddit
It's against their ToS unless you're contributor to open source. If you are not, you are subject to ToS violation.
btvn@reddit
If the problem with GitHub is availability - I'm not sure Codeberg is really an improvement in that area.
IgnoreAllPrevInstr@reddit
Codeberg. I've also looked a bit at tangled.org, where you self host your own node, but it gets tied into a single network, so it all looks like one app
kamatsu@reddit
(you don't have to self host your own node, but you can)
tanaciousp@reddit
Surprised to see sourcehut.org not mentioned here. Never used it but people on hacker news like Drew’s blog posts.
headinthesky@reddit
I've been looking at gitea
Ok_Fault_5684@reddit
I've seen https://forgejo.org/ around quite a bit
medzernik@reddit
sourcehut. its amazing
trannus_aran@reddit
Codeberg plus a cheap VM running forgejo actions, never looked back
miversen33@reddit
Recently moved to self hosted Forgejo. It's fucking slick. Still waiting on federation support but for my own shit, it's great. I still am on Github because it's basically social media for developers. But for my own projects, I host them locally there
chazzeromus@reddit
go hardcore, push to a flash drive
Individual-Praline20@reddit
Ah, they provide exe now instead of code 🤷🤭
TrashConvo@reddit
Despite what they might think, GitHub cant be the hub for agentic coding workflows if they cant get the basics of being a git server right. People will leave if the product is trash and the alternatives are just as good
thewormbird@reddit
My personal usage patterns of GitHub haven’t changed since fully migrating away from SVN nearly 15 years ago. Though I’ve been slowly migrating to my own gitea instance more recently.
Old_County5271@reddit
That's great and all, but personal websites always go offline after 5 years or so, always keep a mirror and I guess push --all if you can.
thewormbird@reddit
That is smart. The git data does not matter as much to me as the files. I don't really do any retrospection of git commit logs beyond resolving the latest git tangles I get myself into.
Old_County5271@reddit
I did not come up with this, fossil autosyncs by default, so keeping multiple remotes/backups is second nature.
Top-Rub-4670@reddit
Seriously, keep a github mirror. Historically, all personal websites and self-hosted things go down within a few years (usually because of a loss of interest/life events, and no, reader, you won't be different even though you're all hyped about self hosting right now).
Github will still be there, in one shape or another. Keep a read-only mirror of all your FOSS projects there. Write in bold that this is a mirror and try to convince them to contribute to your self-hosted instance instead (they won't).
thewormbird@reddit
I've held onto the same 2 or 3 domain names for 20+ years. I just need to find something low cost, isn't AWS, and isn't going to fold for at least the next decade.
I understand the platform risk, even the platforms I host and manage myself.
adnanclyde@reddit
I personally have a daily borgbase backup that I download and check twice a year.
I'm not even hyped about self-hosting. It's just cheaper than 12 different cloud services that are potentially stealing all my IP because it looks like copyright law means nothing if it's for your AI empire.
Old_County5271@reddit
I would agree... except right now what we're seeing is the death of github, if you can't even trust a merge, then it is worse than unusable. If this was the 70's, this post would be titled "Github considered harmful" and it would be 100% right.
There's only 1 website I trust to outlive me and everything, and that's sdf.org
phillipcarter2@reddit
It's fun to poke fun, but there's a world of difference between being a git server for a few codebases and being the preferred, free service for the entire world. Their infrastructure was not built for the amount of traffic they're getting hammered with, and we're all paying the price by tying our stuff up to their services. I guess in my case I don't pay, so I'm not mad, but if I did pay I would be.
tav_stuff@reddit
And yet it worked flawlessly up until they started spamming us with this clanker nonsense
YaLlegaHiperhumor@reddit
No it didn't. It's had uptime problems since at least MSFT's adquisition
phillipcarter2@reddit
It very much did not, and their massive user and org growth since the Microsoft acquisition, not to mention forced migration of various services from AWS and self hosting to Azure, were also contributors.
It’s also important not to have rose colored glasses here. GitHub has always been a home of many flaws in its different eras.
tav_stuff@reddit
Believe me I don’t have rose tinted glasses, but I also remember never actually being annoyed with the state of things until this last year. This is the first time in my life that my literal job is being impeded by GitHub not working
phillipcarter2@reddit
I fondly-not-fondly recall 2016 being that for me. My team (.NET team at Microsoft, actually!) was all in on using GH for development and every day was a crapshoot of if a PR would load because we had the audacity to leave thorough reviews. We had them on the phone a lot and the team was responsive but our “lots of people leave lots of thorough reviews” workflow broke most of their architectural assumptions at the time. It took about a year to get reliable, but we persisted with it and supplement med with some home-grown tools as needed. I don’t think they’d ever dealt with such a density of activity surrounding the code at the scale we operated at (over 300 engineers), so the current times are just rhyming for me.
lurker_in_spirit@reddit
Correlation, not causation, but...
https://damrnelson.github.io/github-historical-uptime/
Darkagent1@reddit
Ehhh I wouldnt put too much stock into a site like that.
https://github.com/DaMrNelson/github-historical-uptime/issues/2
It defaults to 100% uptime, even if there is no data being fed in, so if they started tracking using that page in 2019 then everything before it would be by default 100%
https://www.githubstatus.com/uptime?page=3000
Including apparently 100% uptime in 1996! 10 years before the site was even created.
RoburexButBetter@reddit
It's not about the amount of traffic, we just had to migrate from bitbucket to GitHub and it's atrocious how bad the user experience is, bitbucket gives you a nice overview of your PRs in progress and what to review
For GitHub one of our DevOps guys had to vibecode a greasemonkey plugin to do that, though you can get a chatgpt interface to ask it what you still need to review
Absolutely bonkers
Leliana403@reddit
Imagine calling GitHub atrocious while simultaneously praising Atlassian products.
phillipcarter2@reddit
That has nothing to do with the reliability problems causing ghostty to leave.
needmoresynths@reddit
Tbf they are pushing agentic coding very hard so they're partly to blame here
phillipcarter2@reddit
They certainly are! Especially since things like /fleet in copilot CLI are literally designed to just swarm commits (each of which kicks off a CI run, etc).
My guess is they estimated they’d have a lot more runway to address things last year, as I have no doubt plenty of people internally knew this could happen. But they didn’t anticipate Claude Code taking off like a rocket last Winter. I’m sure a dozen or two SREs there are saying “i fucking told you so” in their heads every day.
DandyPandy@reddit
As an SRE, I pour one out for my homies
DetectiveOwn6606@reddit
What ? Isn't coding solved based on microslop ceo statement . Why would they struggle with scaling,I am sure they have tons of free azure servers. Didn't microsoft layoffs 10000 employees just now because ai is 10x multiplier. why are they struggling
beyphy@reddit
You're assuming that their service is bad because they're getting hammered with traffic as opposed to something else like merging vibe slop to main.
MDTv_Teka@reddit
It's not like they're not vibe coding their platform right
PaintItPurple@reddit
Poor Microsoft, just an innocent victim of all these big soulless companies pushing AI like, uh, Microsoft.
Caraes_Naur@reddit
They had it right, before Microsoft bought it.
Leliana403@reddit
No they didn't. This is just nostalgia talking. GitHub was pretty much in a feature freeze state going nowhere when Microsoft bought it. I'd argue that if it hadn't been bought, GitHub would not be relevant today. No matter how much the doomsayers would love it to be otherwise, Microsoft saved GitHub and aside from a few well known fuckups, GitHub has consistently improved year on year under Microsoft's ownership.
Sorry not sorry.
seacucumber3000@reddit
You dropped this king 🤡
Leliana403@reddit
Thanks for your valuable and well thought out insight, it has been noted.
inkjod@reddit
Fundamentally, there's nothing wrong with that.
hitchen1@reddit
There is when you have competition beicming more attractive by providing more features. Gitlab would have devoured GitHub if they never progressed.
valarauca14@reddit
They would have gone out of business, they were losing money at an absurd rate
chucker23n@reddit
Yeah, but I think that was a perfectly reasonable line to draw. Microsoft didn’t make it free out of the goodness of their hearts, but for PR bragging rights, and now they have to make the money back elsewhere, in a more convoluted business model.
grauenwolf@reddit
GitHub didn't need CI features. I would rather it be a good source control system then a mediocre everything system.
chucker23n@reddit
Well, GH didn’t even have its own CI then.
somebodddy@reddit
Which is not necessarily a bad thing. Everyone just connected external CIs, and the entire ecosystem didn't try to lock you to GitHub Actions.
GBcrazy@reddit
Honestly, from the CIs I used before, GitHub Actions was a game changer to me.
captain_zavec@reddit
Github actions are certainly easier for me to grok than jenkinsfiles, but that may be at least partially due to familiarity.
The other one I've used quite extensively is gitlab CI though, and IMO that one is much nicer than actions.
dkarlovi@reddit
I think both GL and GH have their advantages.
Coda17@reddit
https://damrnelson.github.io/github-historical-uptime/
nemec@reddit
TBH I think Github simply threw away old status history when they migrated from status.github.com to githubstatus.com in December 2018 (conveniently shortly after Microsoft's acquisition closed).
Doing a Google search for historical Github issues led to an incident on March 2, 2018 which is listed with 100% uptime here.
https://web.archive.org/web/20180307004502/https://status.github.com/messages
Also random clicking around:
That or Microsoft is being far more transparent about outages than Github ever was.
foramperandi@reddit
It's exactly this, although I wouldn't give MS credit for it. This graph actually shows the opposite of what it purports to. GitHub had tons of outages before the Microsoft acquisition, but didn't have the operational maturity to actually handle incidents and statusing in a consistent way. What appears to be more incidents post-2020 is actually an increased internal emphasis on incident communication.
People were making the xkcd "compiling, but GitHub" joke as far back as 2013: https://xcancel.com/petecheslock/status/368036953541058560
x21in2010x@reddit
Just jumping around the wayback machine, there are error messages present on other days too. Each incident has at least a symptom and resolution present. So while you may have a point about not communicating enough detail, the amount of downtime itself seems to have been clearly communicated on the old page.
TehTuringMachine@reddit
While this looks damning, to be fair, there are many other things that happened during this tim3 that could at least be partially to blame for this trend.
Not defending Microsoft here, but this is an over-simplification at the very least. For example, most of the real activity in this graph happens starting in 2020 (covid times)
Leliana403@reddit
And also if you look at the breakdown you'll realise the vast majority of the downtime was GHA. Everything essential like core git operations and issues are still pretty solid.
mughinn@reddit
I mean, sure.
Also, 5 days ago they fucked them up for a few hours by absolutely breaking PRs https://www.githubstatus.com/incidents/zsg1lk7w13cf
Witless-One@reddit
No, literally everything is worse. Why are you lying?
tpolakov1@reddit
Because the graph literally says that core git features are stable. Or are we now calling the data fake?
Witless-One@reddit
I went into breakdown and unchecked everything except git operations and it’s worse after the acquisition
phillipcarter2@reddit
Most of all was that after the Microsoft acquisition their growth really started to take off, and Microsoft pushed tons of enterprises to use GitHub over TFS and Azure DevOps. Just an endless stream of growth and scale across every dimension imaginable, now accelerated since everyone and their mother is letting Claude push code at scale.
Twirrim@reddit
I can't help but think they're close to breaching the trust thermocline.
https://every.to/p/breaching-the-trust-thermocline-is-the-biggest-hidden-risk-in-business
unapologeticjerk@reddit
Thank god they figured out Actions and definitely aren't regressing.
chicknfly@reddit
But GitHub invented Copilot. Surely….. yeah, you right.
wise_young_man@reddit
They didn’t invent agentic. They made an autocomplete which was the original copilot. Don’t rewrite history.
grauenwolf@reddit
Reading is hard. Let's go shopping instead!
chicknfly@reddit
I was gonna let the downvotes do the talking 😂
Maybe-monad@reddit
Arrows can't talk, only whistle
iris700@reddit
Sure can be the hub of agentic coding failures though
Spleeeee@reddit
How do they not git it right? Isn’t being down 10% of the time in the SLA?
Sn34kyMofo@reddit
GitHub gets Microslop'd again...
gene_wood@reddit
Here's a data visualization of what's going on : https://damrnelson.github.io/github-historical-uptime/
mexicocitibluez@reddit
There's no way it had 100% uptime across 2 entire years.
pfc-anon@reddit
I don't believe this data, anecdotally it feels way worse.
If it's just tracking the status page, then that's not realistic as that page is manually updated once an incident it confirmed.
dvhh@reddit
plus the recent incident is not impacting uptime
doterobcn@reddit
Has he paid anything to github for this 18 years of entertainment and support?
tav_stuff@reddit
Github is a commercial product run by a multibillion dollar company
doterobcn@reddit
This does not answer my question.
I see a lot of people that complain about services or their quality and they use free tiers and do not pay anything yet expect 24/7 support.
tav_stuff@reddit
Literally nobody here is expecting 24/7 support. Just a product that works
doterobcn@reddit
So he wasn't paying for a product yet expected premium features?. Giving your data away is part of not paying for something i guess.
I do not expect enterprise grade service for free solutions.
tav_stuff@reddit
The product actually working is a premium feature? LOL
doterobcn@reddit
They are under no obligation of having an uptime of 99.9% for non paying customers.
You might like it or not, but it is what it is.
If you pay and get shit service, then you can complain and escalate, otherwise you just accept your fate.
EveryQuantityEver@reddit
Again, it's Microsoft. They have no excuse for not being able to provide that for all of their users.
tav_stuff@reddit
I am not even asking for 99.9%, I am just asking to be able to use the fucking website. Nowadays when I try to use the site, it’s a 50/50 on if it even works. Not to mention that my employer DOES pay, and we still can’t even use the site half the time
doterobcn@reddit
I haven't had an issue in months, probably not a heavy user, but I do use enterprise repositories with no troubles.
EveryQuantityEver@reddit
Quite frankly, they are owned by Microsoft, who has the resources to make sure that Github does not have these outages
mr-figs@reddit
sr.hut everybody
Top-Rub-4670@reddit
Drew Devault, the man who enjoys sexualizing pre-teen girls, generously offering antiquated-by-design git hosting. Versus Microsoft, the company that will rape you and then lawyer-talk into agreeing that you had it coming, providing an increasingly bloated but good web interface.
Tough choice.
Plank_With_A_Nail_In@reddit
Oh no someone I never heard of is leaving.
GitHub usage is mostly about marketing and customer familiarity, people going to think twice about downloading software from a service they have never heard of.
lottspot@reddit
Is it coincidence that GitHub's reliability has become progressively worse as agentic coding has become increasingly prolific? An exercise for the reader.
waterkip@reddit
Yay! People are leaving that horrid place.
DowntownBake8289@reddit
What's horrid about it?
brown-man-sam@reddit
….did you read the article at all?
DowntownBake8289@reddit
Yeah, I did. Read my other post here. I explicitly said what the article was about, and he's not leaving because GitHub is horrid. Did YOU read the article?
brown-man-sam@reddit
Idk man, constant outages that are blocking work sounds pretty horrid for critical infrastructure in CI/CD.
When you hit your nth day in a row of workflows being crippled because of bad service health, it becomes pretty hard to view things in a positive light.
They also expressed that they’re angry at the state of things and have been very vocal and mean about it. Calling it “horrid” in its current state wouldn’t be an inaccurate statement.
And no, I haven’t read your other posts here, there’s 200+ comments and I don’t always look at usernames. Get over yourself.
waterkip@reddit
Its MS owned, its been horrible from that moment onward. And people seem to think git is github or github is git. I
wildjokers@reddit
Seems to have gotten better since MS acquired it. Can you give examples of how it is worse?
waterkip@reddit
I think they don't do it anymore, but they upped subscription prices back in the day and limitted repos.
I already had my gitlab account so I moved to gitlab. Im also using codeberg.
Then there is the issue: github is closed source, microsoft owned. And MS cannot be trusted with FOSS. The fact that MS owns it, is a determent to the product in itself.
-Cacique@reddit
Microsoft messing up a lot of services
hm9408@reddit
Windows, GitHub, what else?
Thundechile@reddit
Bing and Teams.
dvhh@reddit
They were already terrible to begin with
hm9408@reddit
Yeah, they were enshittified from day 0 tbf
gex80@reddit
Maybe I'm out of the loop. What's wrong with Github exactly? I don't use it for git actions because it never appealed to me. But for code repository outside of I think 2 maybe 3 noticeable outages this year, it's been good to us.
We use Jenkins as our build platform.
mikeymop@reddit
Microsoft owns it and is slowly devolving it into an unreliable mess.
It was moved to react which made it very slow to load. Taking 5-10s to open a PR page.
And Actions has an outage on every day that ends in Y now.
Its become a shell of its former self. And now its doung an "opt-out of training our AI against your code"
prone-to-drift@reddit
So that was it, wasn't it .. i kept wondering if my laptop had slowed down drastically or was my memory failing me but GH pages used to load crisply. Now, the structure loads, some animations play out and then the data eventually renders. It's so irritating.
phillipcarter2@reddit
They've been having a particularly bad string of outages and general reliability problems since agenting coding really took off late last year. Far more than normal, and it's seeming like there isn't an end in sight right now, since "by design" behavior (like pull requests kicking off tons of work) are what are being stressed.
d70@reddit
https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion
It’s a result of the Microsoft acquisition and then moving from AWS to Azure. When the foundation is cracked, everything that’s built on top of it is not stable.
OkayTHISIsEpicMeme@reddit
I prefer AWS over Azure but this has nothing to do with GitHub’s woes
foramperandi@reddit
GitHub was never on AWS in any meaningful way.
phillipcarter2@reddit
They were for a while, but that was long resolved. The bigger issue was their own self hosted infrastructure and piles of code optimizing for that specific infrastructure.
TankorSmash@reddit
I think its that there are 100x more commits being made by autonomous agents stressing the system more than anything else
penguinmandude@reddit
This is is it. They’re usage has grown exponentially and they’re struggling to deal with the scale
Leliana403@reddit
If only they hadn't been forced to jump on the AI bandwagon. /s
TankorSmash@reddit
Are you saying this in reply to a comment, or is this a hypothetical counterpoint?
bonerfleximus@reddit
Bill gates is personally hacking GitHub. My friend told me.
valarauca14@reddit
this article is fan fiction
new-chris@reddit
Total bs
Status-Artichoke-755@reddit
There's an article linked...
NowImAllSet@reddit
I mean...did you read the article? The author pretty clearly outlines what's wrong with it.
That said, if you want empirical evidence, check out the unofficial outage metrics page. 87.5% availability for the platform, with most days having at least some type of outage.
Cachesmr@reddit
They've recently broke a bunch of PRs by merging them with the wrong history. The CI workers are also really bad.
robhaswell@reddit
The GHA runners are atrocious. Take the time to set up your own runners.
Cachesmr@reddit
I run CI on self hosted woodpecker nowadays. I agree with you, they are trash
Status-Artichoke-755@reddit
Since Microsoft bought them, they have been unreliable. Their business decisions and practices are highly questionable and people want nothing to do with it
teknikly-correct@reddit
It's all about github actions really - I find it amusing that to most people github actions is a huge part of github, meanwhile we're over here happily using the baseline source control features!
tbh I can't imagine mixing CI with my git provider, simply because I want my git provider to do one thing and do it really well - git it?
Lucas_F_A@reddit
A ton of outages with github actions, timing reliability for github actions (Zig developers mentioned this), some more outages not related to actions, like the ones you've thought of, the recent (a month ago I believe) problem where merge queues deleted work.
awesomeAMP@reddit
Same here, as a code repository only its been great and I like it. We keep our pipelines on AWS because I personally do not enjoy GH Actions.
gajop@reddit
Yup, same. We use GHA with self hosted runners. A few hiccups here and there, but generally smooth sailing..
deadcatdidntbounce@reddit
Microsoft are rehosting GitHub on Windows 11 for a better user experience. Are you enjoying it yet?
Thundechile@reddit
Also moved all of my personal projects away from Github, the way they decided to do AI training on people's code (you have to specifically opt-out) was too much of a dick move.
XTCaddict@reddit
Wait what
lngns@reddit
GitHub ToS have you give them a licence to your code for any purpose and have you waive your right to sue them for it.
The ToS always were like this, but people did not like it when GitHub starting using for AI training.
pjmlp@reddit
I would bet most folks responsible for creating it, are no longer under Microsoft paychecks.
This is what happens with most acquisitions.
Xamarin one, also went down quite bad I would say. The only thing left of it, is the infrastructure used to target iOS, Android and WebAssembly. Everything else was either replaced by modern .NET, or rewriten in incompatible way (Xamarin.Forms => MAUI).
bzbub2@reddit
this post is like, way to raw and emotional for what seems like something that is just such a bandwagon right now "oh no github has 2 nines availability lololol". it isn't causing me trouble personall so maybe i'm missing the point but it's like, why get so upset when you can't hit refresh there sometimes. are their release workflows failing? maybe that matters. otherwise, barely registers
davidcelis@reddit (OP)
yeah! emotions are dumb and should be hated!!!
tav_stuff@reddit
Dude he’s getting emotional over a fucking git forge lmao
malakhi@reddit
He says it in the article. It’s been hampering their releases, their review process, everything.
bzbub2@reddit
Ctrl+f release (0 results) Ctrl+f review (1 result, yes, he couldn't review a PR for 2 hours) description of him writing an X on a physical calendar, like he is a prisoner in a virtual github prison, counting down the days to his inevitable death...yes there is one result for that as well
fygy1O@reddit
What are some alternatives that people use?
tav_stuff@reddit
Codeberg
Practical-Positive34@reddit
What a moron lol...Good luck to him...I think Ghostty days are numbered anyways. I could literally recreate their platform in a matter of a month. Maybe I will just to prove my point.
ferric021@reddit
My biggest takeaway is Mitchell being less than 36 years old as accomplished and prominent as he is.
Krigrim@reddit
Ive had a lot of issues with GitHub actions as well so I can’t blame him. Been thinking about going over to Gitlab instead
_BreakingGood_@reddit
Every issue in GitHub is, 10x worse in gitlab. I thought I hated GitHub until I joined a company that uses Gitlab.
zsaleeba@reddit
I've used gitlab for years and it works great for me
Witless-One@reddit
Have you used GitHub in earnest? Please be honest. I’ve used both extensively and anything Gitlab does GitHub does better. And there are many things that GitHub has which Gitlab does not.
Gabelschlecker@reddit
Do you have some concrete examples?
Cory123125@reddit
Can we be honest and say that no one wants to go making a detailed case for any reddit comment and all of us have vague opinions based on past experiences that we are not going to specifically dig up?
TankorSmash@reddit
Fair enough, but it's also way more convincing if you provide examples. One thing to vent, but if you're trying to change someone's mind, I don't think saying 'trust me bro' is enough!
Cory123125@reddit
Of course. It's just that its a big requirement for casual conversation.
Zweedish@reddit
I've used both extensively (way more gitlab though) and Gitlab is significantly better than GitHub in most ways.
XTCaddict@reddit
Well that’s just not true CI is a breeze on GitLab it’s largely automated
KawaiiNeko-@reddit
Could you provide some examples? Genuinely curious. I've been using GitHub for nearly a decade now and have just recently started using Gitlab for some particular things and it's been a breath of fresh air.
_BreakingGood_@reddit
It's hard to list them from memory. They're things you notice as you use it.
Some examples are aggressive pagination on the "changes" tab of an MR. Aggressive collapsing of "large files" on the changes tab. Want to ctrl+f a specific string to see if it exists in the changes? Sorry, you can't because that change is on page 2 or in a collapsed file. Ok then let me open it in the "Web View" so I can see the full MR changes and search there. Oops, Web View doesn't support search yet. Meaning, there is literally no way to ctrl+f a string in an MR without cloning the branch locally and searching locally.
Linking directly to a line of a file fails at least 50% of the time.
The worst thing in my opinion is that MRs will sometimes open to a seemingly random specific commit in the MR with absolutely no visual indication that it did other than noticing it in the URL. When this happens, you may not realize you're only reviewing 1 commit from the MR and not the entire MR.
The revert button on MRs exists but is needlessly convoluted. In GitHub, you click "Revert" and it opens a new PR with the exact opposite of the changes in the MR against the branch you merged it into. In Gitlab, it has an incredibly convulted poorly explained flow that I usually just fumble through randomly when the reality is that I want it to do what GitHub does 100% of the time.
GitHub also includes the merge commit when you select "Squash & Merge". Gitlab does not. So the main branch has completely polluted history of pointless merge commits.
I think these are the main things. But they're issues with the core, basic functionality that I experience every day. Not weird edge case issues.
Jay_D826@reddit
I use Gitlab for school and it’s been pretty decent so far. I use my GitHub account for personal stuff and I’m way more familiar with it but I’m ready to jump ship as well. It just sucks that private equity or big tech companies buy up all of these genuinely good and useful services and turn them to shit.
Like, we can go to gitilab or whatever other alternative but if it gets popular enough it’s just going to be the same thing all over again.
this_knee@reddit
I must be more outta the loop than I realize. I know what vagrant is I don’t know who Ghostty is.
killver@reddit
An overhyped terminal emulator
Fenzik@reddit
Ghostty is a lovely terminal emulator by, as it turns out, the same author as vagrant. I’ve been using it for just a few weeks but it’s very nice.
Ok_Slide4905@reddit
Who?
youngbull@reddit
For fun, I have been toying with the idea of seeing what it would be like to use a minimal self-hosted setup. You can sort of get ci by simply using a post-receive hook, then you just display the logs. Once you have that, you can manage VMs by pushing Ansible playbooks. Now you have pretty much bootstrapped a infra-as-code setup and can have it host whatever you like, like your app or whatever you need for development (bug tracker etc.)
I think now, the viability of such a setup is real, compared to something like gittea.
kaeshiwaza@reddit
This setup just works since decades...
scoobybejesus@reddit
No one so far has assumed this is because of the recent issue where the PR being merged ended up being put on a different commit, thus git history being erased and potentially quite a bit of time to untangle the mess. Having the UI telling you one thing and then merging with a random prior commit is a bad look.
ScottContini@reddit
Link?
BrenekH@reddit
The merge problem probably didn't help, but the article is pretty focused on downtime. It also mentions in the footnotes that they've been considering and planning to move for months.
juankman@reddit
People need to remember Microslop is behind this. They shot themselves in the foot with their push of poor quality products.
MateTheNate@reddit
Part of it may also be due to the enormous amounts of vibe coded crap DOSing the service as well
RaccoonElaborate@reddit
I wish I had anything in my life that I cared about as much as this author cares about Github :[
goranlepuz@reddit
I love the smell of ~~the napalm~~exaggeration in the morning.
Immediate-Sink-4382@reddit
Recent outages and the constant CLI/API hiccups make it hard to trust them with anything beyond vanilla git. How's agentic coding supposed to work on a shaky foundation?
SrMortron@reddit
who?
radarthreat@reddit
Just the guy that created Terraform. What have you done?
SrMortron@reddit
No idea what that is.
radarthreat@reddit
I’ve learned to Google, which apparently escapes you 🤣
SrMortron@reddit
Why google something I have no interest in?
ZjY5MjFk@reddit
If you don't care enough to google, why do you care enough to ask on reddit and then get into arguments with randos? lol
SrMortron@reddit
Because I can.
neuronexmachina@reddit
Oof:
betazoid_one@reddit
So he’s creating a new git web app?
Intelligent-Use177@reddit
Github is going to lose for becoming bloated
r2vcap@reddit
I get why people are annoyed with GitHub, especially after the outages. But GitHub is still where everyone is. For FOSS, it’s still the default place where things happen.
AI slop is real, but moving won’t magically fix it. If another forge gets popular, the same garbage will show up there too.
Lock down PRs, restrict comments, require approval, whatever. That seems less painful than moving everyone elsewhere. Just please don’t pick GitLab :( It’s slow enough that I often give up before contributing.
ShacoinaBox@reddit
well, I guess when u put it like that, nothing will ever change and everyone will always stay on gh forever. in fact, im posting this comment on slashdot at this very moment!!
Agent7619@reddit
Makes you wonder how we ever got off SVN
v4ss42@reddit
None of that prevents GitHub’s own theft of IP to train their bullshit generator.
ArkBirdFTW@reddit
git itself isn’t ready for the new paradigm
Status-Artichoke-755@reddit
I moved my projects to gitlab last month. I imagine there will be a mass migration over the next year
SpectralCoding@reddit
Who cares? Throwing out a few hours of lost productivity here and there to just dedicate dozens and dozens of hours and money running your own solution. This is like the bootcamp CTO guy getting all that press moving away from the cloud a few years ago.
cgoldberg@reddit
Apparently YOU care? Also, nowhere on the article did he mention doing it "in house"... and as a public open source project, I doubt he is buying anything anyway. Why are you so outraged that a project wants to move to a new hosting service after frustration with the existing one?
bonkyandthebeatman@reddit
did you read the article?
SpectralCoding@reddit
Yes so instead I hope he will write a “-“ on every day he spent working on whatever new solution he’s moving to.
citramonk@reddit
There are existing solutions, migration is usually documented. Not like they’re gonna create a new GitHub
SpectralCoding@reddit
Just track the engineering effort to migrate and refactor workflows and permissions. I hope he finds the effort to migrate and the new provider’s availability is an enabler over the GitHub outages.
Jason3211@reddit
Popping in to say the article was clear, concise, and honest. As I'm getting a bit older, I notice how much I enjoy honest writing with restrained brevity. It was a lovely read, well done (and frankly, the issues at GitHub are maddening to anyone touching repos hourly or daily)!
ciemnymetal@reddit
Reddit and bitching without reading the article, name a better duo
axonxorz@reddit
They just wanted to soapbox and feel superior. Wrote a whole paragraph to show us how little they cared.
Dunge@reddit
Who or what is Ghostty and why does their opinion matter?
Alchemista@reddit
Ever heard of terraform?
remy_porter@reddit
Ghostty is a popular terminal emulator. The developer behind it has been involved in many successful projects.
dom_ding_dong@reddit
Ish.
Xterm ftw always!
/s
Use what you must
alizardguy@reddit
The amount of bullshit Github has put me through makes me very willing to use literally anything else, I'd prefer managing 100 logins for Git forges than use it atp
okilydokilyTiger@reddit
Doesn’t mention where they are moving to interestingly
helloworldpi@reddit
codeberg if I had to guess. Zig already made the move there.
radarthreat@reddit
Mitchell is a big Zig fan so wouldn’t be surprised if he follows them
damesca@reddit
Right... but it does specifically say they're still working out where to go to.
Prudent-String4749@reddit
Even free users get hit when the whole world's repo depends on their uptime. It's a fragile monoculture whether you're paying or not.
space_prostitute@reddit
K.
DowntownBake8289@reddit
He's complaining about the outages at GitHub, the thing that people are complaining about right now. If GitHub fixes that, he'll be right back.
radarthreat@reddit
I mean, he talks about how much GitHub has meant to him, and he and Armon hired their CEO from GitHub, so yeah, I’m sure he would be back if it worked properly.
thepurpleproject@reddit
They need to first separate their infra from paying customers and free slip machines. Then have a consistent pattern of achieving a thing. It seems GitHub problem is multi layered. Actions, breaking UI, backend shots on large prs, massive artifacts and their whole wip ports to react
CoronaMcFarm@reddit
Microsoft is unable to do that, what you are asking is impossible
mikeymop@reddit
That's two I've read today.
Bookstack moved to Codeberg and setup their own mirror.
Steinarthor@reddit
Why doesn't Github tell CoPilot to go fix their problems...are they stupid or something???
markus_obsidian@reddit
Good for them.
Unfortunately, my org is entrenched. We put all our eggs in the same basket, because why wouldn't we? The cost to move our CI elsewhere is staggering.
Github is banking on sunk cost.
watabby@reddit
Even with the company I work at the outages have had a measurable impact on “developer productivity” isn’t something that’s explicitly measured.
We’ve even had to delay client onboardings a day here and there. Something we can’t afford to do considering that we’re a startup and any lost contracts would be devastating.
awmath@reddit
Any production repos I and my company have are not on GitHub. And that's perfect. Only open source projects end up there. Usually for the exposure. But AI has pretty much destroyed all GitHub usability.
Looking for a solution to a specific problem? Good luck with thousands of vibe coded projects with a single commit 3 month ago. Do you have an open source project on GitHub? Have fun with bot generated PRs completely unaligned with the projects vision.
I can absolutely understand the motive and I wish the project the best of luck.
brutal_seizure@reddit
Wow, want some cheese with that whine?
FaradayPhantom@reddit
🤷♂️