I'd argue that the risk of your own instance being inaccessible for whatever reason is significantly higher than GitHub or GitLab having a full outage.
I think the predominant factor is data governance and information privacy. A self-hosted GitLab instance is mostly good enough, until you don't want to self-host it anymore and change to a cloud hosting service of your choice. It's just that not a lot of privacy aware people want to pay Microsoft for playing around with their data.
In all of the companies I worked, the "basic" tools (aka mostly GitHub, Google Workspace, Jira, Slack etc …) were all just default choices from the era when the company was fund and then those things just became the base bricks of thousands of internal procedures and scripts and once the company is at cruise speed, nobody will want to change any of those tools even if "better" (for whatever definition) or cheaper tools exists.
Because it’s just too hard.
And somehow, by some amazing magic I don’t understand, at some point, people are happy with the fragile stack of hacked things running on top of 4 or 5 extremely "shitty" (for whatever definition) products that you can’t cancel.
I won't disagree with your assessment that the tech stack changes rarely or ever, but newly founded companies are still picking GitLab fairly often. AFAICS every business wants their tech stack "moved to the cloud", but infra people weight the costs and still have good contacts to local data centers from the pre-cloud-era - you know, contacts with a phone numbers and human support. Those data centers made it a business to run GitLab instances for you, among other things. A lot of businesses choose a discounter like Hetzner first and then move to one of those when support matters.
That is mostly the best overall price (direct and indicrect costs) for the offered value, data governance included.
They don't want to invest in integrating their product in Gitlab, because it's 1/4 the size of Github and they see trends pointing to a Github resurgence.
If you have limited resources to build integrations then sure, I guess?
We don’t support Linux because the low marketshare on sales.
And that why a lot of people suffer from bad software - it was the first known and they don’t think in long term.
PS: GitHub is usually not an option for closed-source software. Not because of the uncertain development of the fees. Microsoft’s security track - especially for the cloud - is bad. And if you lock yourself to cloud services you end up in vendor lock-in. And that’s also open-source projects tend to Gitlab and others. As usual, GitHub was the first big player.
This is like freemium^2. People aren't buying gitlab licenses because a) too expensive [1] and b) not needed for the functionality in the free product.
[1] "too expensive" is in the eye of the beholder and requires visibility of ROI, COGS, and development velocity.
I never understood that attitude, but I'm not the one who makes financial decision and tinkering with stuff is somewhat fun for me, so I don't really complain.
Do you distrust the new for-profit company that owns Gitea? Do you believe that all software should be libre? Choose Forgejo.
Do you believe having a for-profit company behind gives the project more budget for feature development? Do you think that using a GPLv3 licensed software is problematic in legal terms? Choose Gitea.
Yes: https://docs.gitea.com/usage/actions/overview
> pages
No, but if you have some sort of static content hosting set up (like a S3 bucket), it shouldn't be difficult to set up publishing to that with actions. It's also got project wikis built in.
> user permission management
The only thing I can really complain about are the constant UI changes that do not actually improve anything. I fell they are just changes for change's sake to keep some frontend devs busy.
Employee salary is fixed expense your company is always prepared to pay every month.
Buying new software is extra new cost, which needs justification and procurement process.
Side comment to the screenshotted tweet. You can like open source with out making commits. It's nice to just be able to read the code or take something and just run it or fork it privately
No real criticisms of Gitlab itself, just someone being weirdly upset about being asked to support more than 1 git hosting platform because some people don't want to buy into Microsoft's ecosystem and trying to argue that you should just use Github purely because it's more popular and more profitable, with the obligatory ugly AI-generated image at the top.
In particular after the US election - Github is owned by Microsoft and they love training their AI on your code (which i don't mind for open source projects).
The "no costs" is a wild argument to me though, because I've seen "we want it locally available" and "we can't put code in the cloud/"laws"" etc, but never "this is cheaper than a couple bucks per month"
I think GitLab having a quarter of the revenue of GitHub is incredibly impressive, and makes me take GitLab a whole lot more seriously.
- Github: 2 bil. revenue on 6K employees = 333K/employee
- Gitlab: 547 mil. revenue on 2300 employees = 237K/employee revenue
interesting, github is more efficient on that metric ... scale matters I susppose
And the final paragraph is the kicker: "I’ve nothing against GitLab, and their software might be pretty good. All we know is that they’re outsiders, while a lot of the tech market in Europe seems to think that betting on GitLab is the best go-to-market strategy."
Sounds like they haven't even evaluated gitlab? Talk about the pot calling the kettle black. At least I've heard of Gitlab! If your advice is not to cooperate with outsiders or to even evaluate their offering, well... The French have amazing marketers, but they suck at writing coherent blog posts ;)
Regarding market share of GitLab, I would also assume that it's not looking good, but I only have anecdotal evidence. At least we (350 premium seat license) are definitely looking to migrate away from GitLab, for various reasons, but mostly cost. It seems the Premium tier has become the ugly step child of GitLab, but Ultimate is simply way too big a step cost-wise. Also, the new AI features have a hefty additional cost, and at least for coding, Copilot is superior in my experience. Combined with the overall bugginess of GitLab and missing essential features like proper code search (yes, I know, they will ship with Zoekt any day now), it's in a bad spot now compared to GitHub.
EDIT: I forgot one very important additional thing: GitLab was attractive in the EU for regulatory reasons, as GitHub was US hosting only. But since end of October, GitHub has EU hosting generally available:
https://github.blog/changelog/2024-10-29-github-enterprise-c...
Or has this changed somewhere in the last 4 years since I looked at it?
See perhaps:
* https://en.wikipedia.org/wiki/International_Safe_Harbor_Priv... (expired)
* https://en.wikipedia.org/wiki/EU–US_Privacy_Shield (expired(?))
> The EU parliament raised substantial doubts that the new agreement reached by Ursula von der Leyen is actually conform with EU laws, as it still does not sufficiently protect EU citizens from US mass surveillance and severely fails to enforce basic human digital rights in the EU.[7] In May 2023 a resolution on this matter passed the EU parliament with 306 votes in favor and only 27 against, but so far has stayed without consequences.[8] The NGO NOYB (European Center for Digital Rights) has announced that it will once again try to set the Framework out of force in front of the European Court of Justice.[9]
* https://en.wikipedia.org/wiki/EU–US_Data_Privacy_Framework (in place)
Running your own Gitlab instance or your own instance of anything really is by definition more upfront expensive but also a hedge against future costs. Cloud deployment is one example of it. There's been a pretty significant move back to on premise because that's the exact dynamic that played out.
If there's a French cultural element to it it's not "people in France can't calculate costs", which is both patronizing and obviously stupid, it's that the French have a healthy skepticism of outsourcing their infrastructure to people whose decisions they have no control over.
Due to a large & spread code base (~1000 repos) and rather complex testing and deployment needs (we sell research robots: many different platforms, each of them with customer customizations), Gitlab has been an excellent platform for us, that we can easily shape to the needs of our pipeline.
The dev experience is also great: pushing one tag essentially turns code in ready-to-install Debian packages in a test environment, followed by one additional click on a Gitlab pipeline to send the packages to production.
We could probably do that with github as well, but I would not underestimate the convenience of having the whole platform accessible and available at our fingertips.
TL;DR: we won't move to github any time soon :-)
Not implementing everything customers ask and focusing on supporting only one platform/forge and do it very well is very respectable, but I don't feel the explanation given by the author of the article is very convincing.
I scanned their homepage briefly and it seems that Gitlab already does most of the stuff they mention, including the AI stuff that's so popular right now. "CI Issues revolutionizes the way teams handle CI problems by leveraging the power of AI to automatically detect and surface issues." Yeah, ever heard of the /summarize command?
So in conclusion, it seems to me that if people want to pay for something to improve their Gitlab experience, the money would be better spent on a Gitlab licence to unlock additional features.
Ironically, GitLab auth is currently broken. I try to sign in via GitHub and am served a 422 Auth Error page. Perhaps they don't have many users that use GitHub to sign in, or perhaps the sign in page isn't tested? Whatever the case, the first button I press on their website being broken is exactly the same feeling I remember when I tried GitLab for the first time about 8 years ago.
I think you are mixing up Gitlab and Github?
This is an absolutely wild statement to me. I find github actions to be really terrible for any reasonably complex CI flows; gitlab's CI is the primary reason we use it instead of github. If github's CI was worth a damn we'd certainly be there; as it is we mirror a bunch of repos to github just for discoverability.
In any case, GitLab still has some advantages over GitHub. DX, UI and UX of GitLab CI is better than GitHub Actions. And one massive one - you can actually have a complex organisational structure without weird ramifications. In GitHub you have orgs and projects; in GitLab you have orgs, namespaces (which can be nested!!) and projects, and you can move projects around with automatic redirects. That means that you can actually group projects together, have them share CI configs, secrets, etc. In GitHub if you want to separate or group stuff, you have to create extra orgs, which results in extra SSO and token configurations.
Yeah, GitLab Community self-hosted is pretty popular in France and more widely in Europe. That's because it's good and does its job, and it's actually quite easy to maintain, and it's entirely in your own control (how often are GitHub down? Did they train their LLM on your code? Are you beholden to US foreign policy and sanctions?). Having been responsible for running GitLab for a small to medium sized org with very heavy usage for a few years, I'm extremely convinced we got better ROI and uptime than paying GitHub.
French engineers might go a bit too far on the DIY with free software route, but on the other end American engineers go way too far with "let's just buy another SaaS for this small thing" IMO.
If anything, it's attack on French bureaucratic processes that make it hard for employees to ask for something to be purchased. Or it's a cultural thing, once you're employed you're a fixed cost and your hourly rate is not analyzed anymore in purchasing decisions.
The results are the same. It's easier to spend 10 hours on a free thing than to get approved a purchase for something that costs one hour. Numbers are just an example, don't know how they hit this specific case.
Therefore, when considering market for your software, it's not only about demand and usage, you have to take into account whether that demand comes from those who are willing to spend money. The fact that 50% of developers in France use Gitlab, doesn't mean much if they're not willing to pay. If above the average (compared to other countries) 50% of usage is specifically because of the free self-hosting option, it preselects those who will spend time instead of money.
Anecdotally, sales of my software in France, or for that matter in Spain or Italy, are 10x smaller than sales in Netherlands, or Sweden, much smaller countries. maybe it's a language thing, as some countries are know for excellent levels of English, and some not? But since my software is for developers and IT people English shouldn't be a barrier for them, so I suspect cultural and bureaucratic reasons are the main culprit.
Github is owned by a company with, some might say, in the long run a terrible track record.
Github is starting to get slow and heavy.
By the way, congrats to GitLab for its $579 million revenue and kind of staying in the race. It is not easy to compete with the borg hivemind nowadays...
Fwiw: Gogs, the project gitea forked from is still being maintained and continues to be just plain open source.
It doesn't have full vertical integration though (built-in ci etc)
And in Europe topic of data governance is very much a thing...