Gitlab is reportedly up for sale
developer-tech.com
developer-tech.com
I've also never felt compelled to use Gitlab. They literally offer no significant advantage other than being an alternative to Github.
If I started over, I'd go with Drew's Sourcehut hub for the sake of actually shooting for something different.
I don't really see who would pay $8B for this business which has no moat-future.
I can see how GitHubs self hosting story might be worse, because it’s just not designed for that, most companies don’t do it anymore, and I believe the product is deprecated.
As for doing devops from the platforms, I always found GitHub Actions better than GitLab CI, but I realise many feel the opposite.
While I agree with most of your point, I'd suggest that being self-hosted is a fairly major draw for a certain group of people. There's a lot of overlap with the anti-MS cohort, but it's not exclusive. Most of the uses I've seen of GitLab in companies is those who stick it on a cheap VM and never pay for it.
GitHub has had a self-hosted offering in the past, the current availability is... cloudy, looks discontinued, they're trying to move people back online. It was always way more expensive though, targeting a completely different market to GitLab self-hosting.
[1] https://docs.github.com/en/enterprise-server@3.10/admin/over...
so you use Github? if they're the same, why not use the alternative? I just don't understand your negativity given that you don't list any negatives
Or Codeforge (Forĝejo, if self hosting).
The problem is that SourceHut is not becoming any better over time. With all my respect to Drew, his attention is spread too thin over many projects.
It's not a very useful metric. A user having a flaky internet connection with high latency and low bandwidth and intermittent disconnections would complain that the platforms he uses are "slow".
Or a service is just bogged down by too many users but would be quite responsive if not overloaded.
EDIT: I was not writing about JIRA specifically. I agree that JIRA is ALWAYS slow. Ten years ago I had to wait five seconds after a button click. However I also had to handle a complain that something was slow just because that customer's internet was shit.
EDIT 2: I remember ten years ago an especially perverse variation. A friend of mine was complaining about the bad quality of video even after having upgraded his internet connection. The problem was asymmetric bandwidth. His friend had bad upstream and my friend just saw jerky movements and ugly artifacts and it was unusable for communication in a Signed Language.
I learnt that day: if something is slow, then it is not always easy to understand why. My friend was very disappointed in me.
On the contrary, a bad internet connection makes it much easier to distinguish slow and fast platforms.
Video tends to be bad for the receiver if the sender has bad upstream.
That being said, I haven't seen anything egregious for a couple of years, so maybe he's taken some of the honest feedback to heart.
As for Gitlab and Github, I left ridiculous Jabbascript abominations behind and never looked back; Github now overriding usual keys like PageUp/Down and Home/End SPA style and not being able to display most stuff without JS is the straw that broke the camel's back for me.
Gitlab tries to do a lot and never did anything that well. It's just perpetually catching up to github.
Last time I used it, which was a few years ago now, it felt like Github had caught up and a lot of the Gitlab features were flaky or didn't seem to serve any purpose other than checking a sales tick list. Stuff like integrations that don't seem to really serve any purpose other than to say they are integrated.
I don't think integrations are the example you think it is. The ones I tried work well and make it trivial to setup features that would otherwise feel like papercuts. I'm talking about things like being able to handle gitlab events without having to manage anything or provide your own event-handling infrastructure. If you have web hooks you need invoke or react to, as anyone who runs full CICD and cares about blocked pipelines does, with GitLab you don't need to worry about exposing custom endpoints and write controllers. For many, that is the difference between using notifications or not.
I find this remark baffling by how much it contrasts with reality. Others in this thread already commented on GitHub using terms like "mediocre". I wouldn't go that far but I am indeed aware that GitLab feels it works and is expected to work reliably, specially CICD which, unlike GitHub, they succeeded in turning it into a solved problem.
It's even more baffling reading bold claims like "Gitlab tries to do a lot and never did anything that well" when only a couple of years ago GitHub Actions were renowned for not even being prod-ready,whereas GitHub CICD just works and works well without requiring investing any thought into it at all.
I never had any issues GitHub Actions for CI stuff at my previous two companies, both which used it heavily. GitLab CI also works fine, though I struggle to see how anyone could strongly prefer one or the other as my own experience is that they basically work the same. Both basically boil down to some yaml you can use to configure running stuff in docker containers.
but GitLab kept adding the good stuff that Actions introduced, and it can do a lot of things, and with the Omnibus package it's very easy to self-host. (and upgrades are well supported, etc.) ... and of course GitHub self-hosting is only for enterprise editions.
sure, you can setup one GHA Runner on one VM (and each one one a new one), but that's it. nothing else is supported officially.
I wasn't using GitLab 6 years ago so I can't comment on that, but I can comment that at least since 2022, I haven't found anything in GitLab that would recommend it over GitHub.
I don't think you got the point.
The whole point is that a couple of years ago GitLab was unquestionably the market leader and hands down the best service.
And since then it didn't got worse.
Best case scenario, alternatives like GitHub managed to put together similar offerings. That does not mean GitHub suddenly was the best. Far from it. Again, others in this thread used the term "mediocre" to describe GitHub, and this happens years after GitHub started to try to catch up.
To me, GitLab's CICD is by far the absolute best CICD system around. It has been like this for maybe a decade now. The container-centric approach to build jobs, a domain model for pipelines that is extremely simple and yet misses no usecase at all, UX that's unparalleled to the point that, unlike other CICD systems, doesn't even require a tutorial to get the basics to work... Things are so far apart that it boggles the mind how anyone who has any experience in CICD would even mention GitHub on the same sentence.
GitLab was the absolute best 6 years ago and, in spite of Microsoft's takeover of GitHub and rushing to bridge the gap, it still is. That's the whole point.
self-hosting GHA Runners (plural!) is ... hard (not impossible, and quite doable with enough investment, especially if one takes the dive into the murky waters of k8s), but it's not officially supported. (what's supported is here's a tarball we support these distros ... have fun.)
programming in YAML is still bad, composability is still an afterthought, etc.
> programming in YAML is still bad, composability is still an afterthought, etc.
I disagree on that. GitHub actions are as composable as it gets thanks to the ecosystem of actions. GHA is better than GitLab CI in this regard, and probably the best system I’ve used so far. It IS a bit clunky but overall I’ve been pretty satisfied with it.
I also disagree regarding YAML because you don’t program a CI, it’s mostly declarative work. The only conditions needed are a bit clunky to setup, I give you that, but it encourages to either keep the CI simple (no CI needs overly complex conditions), or move the logic to CI scripts in any scripting language.
It is a terrible experience riddled with many gotchas. I'm making WarpBuild to provide runners (cloud and BYOC on user's aws account) to provide more powerful, faster, and overall much better experience. Try us out - you're guaranteed to save both money and time.
yes, writing programs (ie. an action) is the correct level of abstraction, but in general it just doesn't make much sense to have separate steps. (because for many cases unit testing and building artifacts can/should be done at the same time ... for example python/rust ... inside a container, download deps, build, then run tests, copy result to new layer, push)
I’m the maintainer of RunsOn (https://github.com/runs-on/runs-on), which solves this problem with far less investment
I do wish Gitlab would spend some time better organizing their settings menus, but aside from that it seems to be more complete.
That's not how I remember it. I agree that GitHub is ultimately the more polished product, which is why I eventually went back with my personal projects to it... But most of these platform features like CI Integrations, project repositories etc started at gitlab, and GitHub effectively copied them later
Github may have beaten Gitlab to the punch with Pages (whoopee) but CI is a far more important feature, and Gitlab CI existed for years before GA, and continued to be a far superior solution for a long time (and maybe it still is?).
Also, because Gitlab targeted enterprises they had a bunch of features that Github did not. Self hosting, Identity and role based access, weighted issues, project milestones and burndown charts, private issues, support for "innersourcing" of internal repositories. The list goes. Unlike Pages, which is useful but can be replaced by any other wiki, things like role based access to projects are really important. When I switched jobs a few years ago and had to use Github, it felt more like a hobbyist's web-based toy version, lacking a lot of "big-boy" features.
The one truly essential feature that separated them "for the 99%" was CI. Once Github introduced that, it shrunk Gitlab's lead.
But Gitlab are by no means playing catchup to to Github.
Totally agree on CI being superior. The config syntax is a bit of a pain and is aching for some higher level abstraction. But it’s very powerful, e.g. easily supporting cross project dependencies and build status across project in one neat interface.
I also agree that GitLab was optimized for enterprise and that’s where the major feature disparity is. If you are using it only for small hobby projects you will feel that it’s nothing special. I’ve used a self hosted version at a fortune 100 and it was a huge boon to productivity there over GH EE.
CI/CD components (GA since 17.0) improve the situation a notch: https://docs.gitlab.com/ee/ci/components/
GitHub also has mediocre issue management and a wiki and we have GitHub Actions as well.
You still need to integrate your cloud service provider with GitLab, which is exactly the same hassle as integrating it with GitHub.
Gitlab had many features first. Namely, they had integrated CI way before GitHub did. Their open CI runners are really cool and easy to setup. GitHub had to scramble to put out their Actions product to compete.
I don’t know what the feature comparison looks like today, but many of the innovative features you probably use on GitHub were on Gitlab first. They had a way more innovation velocity for quite a while.
no corp will use sourcehut. That's just outside of the possibility. And corps have the most money.
They have exactly three advantages over Github:
1) Ability to self-host.
2) Open Source core (if you carefully avoid all the proprietary features).
If you need a truly open source solution then that would be Gitea not GitLab.
Can you clarify what you mean by this? I ran a hosted Gitlab instance for a previous company for years with SSO, actions, container registry, on Gitlab CE. I never felt like I was missing anything.
While I'd definitely reach for Gitea if I wanted to be a purist about open source I honestly can't think of a Gitea feature that's not in Gitlab CE.
I completely disagree. GitLab undoubtedly has the best CICD service. GitHub Actions feel poorly designed and half-baked in comparison to the point they barely feel prod-ready.
It's interesting to read into the thought process of anyone picking GitHub vs GitLab, because, other than being the default with tolerable shortcomings beyond copilot, there is no compelling reason to pick GitHub.
If anything, GitLab's Achilles heel is it's massive Product Management problem. I recall that access to some features like Group Access Tokens were unexplainably removed from free access without updating documentation or even it's UI. Apparently some GitLab PM made a call to pull them out without bothering to check for the customer impact loose ends, and all loose ends were simply left in place for customers to repeatedly pester support without any action.
GitHub Actions feels much more well thought out to me. The templating feels less hacky and the workflows via "uses" work really nicely for abstracting common bits of CI config into "packages". The downside is that GitLab CI feels easier to pick up and explain.
On the opposite side, GitLab CI feels like it was made for containers and k8s whereas GitHub Actions feels like it's just a VM (even if the runner software is technically more flexible than that). From a management perspective GitLab is nicer but as a user I often want to avoid thinking about any of the k8s quirks and just have an emphemerial VM to play with.
They literally offer self-hosting at the same price as the SaaS version, which is still significant to some of us.
I'll go rather go back to paper than to just move every last bit of my companies data into the hands of a single company (looking at you, Microsoft).
To me, given he critical importance of a working CICD, it feels like Microsoft controlling both GitHub Actions and Azure Pipelines poses a significant risk of having the proverbial rug pulled from under you. I don't expect Microsoft to invest on GitHub Actions enough to make them the clear winner in UX and cost, for example. Both Azure Pipelines and GitHub Actions are appallingly bad in UX, at least compared to GitLab CICD or even CircleCI or Travis, and I don't expect to turn that around due to their incentives to not make either one too competitive regarding the other in-house service. But that's just me.
The two best CI tools I've ever seen were Gitlab, as its approch was just working good (tm) and Teamcity, as its approach with the UI & Kotlin integration was just nice.
Then I've seen Azure Pipelines, which seem like an unfinished product because it competes with Github Actions, and then there is Github Actions which seem to come from someonw who loves to write k8s yml files (they do what they do, but they're unreadable).
I've used 3 of those but I guess not enough to prefer one over another.
Too bad, I like Drew and am benign (despite the skanky username).
Username: sleaze / sleae
I can self-host the community edition for free, and in addition to use it as a software forge and static websites hosting and deployment platform through GitLab Pages, I can use it to manage my users and have third-party services (like mattermost, hedgedoc pads, and other internally developped apps) for centralized login.
I really really hope that the buyer will keep the Community Edition alive, otherwise we'll have to move everything over to something else (I don't know what yet).
disclaimer: While I'm not directly involved with forgejo, I am part of Codeberg e.V., so consider that to be kind of an ad.
The main one are CI/CD Pipelines fully baked in: This is huge. It's fully integrated in the platform. You have full visibility over the CI/CD pipeline. You can even use their own runners.
If you are trying to introduce source control to a traditional org that haven't been using it ever, then a self-hosted option with CI is a much easier sell than something in the cloud (in my experience).
The best CIs are exactly that. Just a task runner, with some reporting capability. Even the reporting capability Github is not very good at.
In most large orgs I worked for they used Gitlab as their main platform (even though they had uses for GitHub too) for the main reason: they could host it whenever they wanted it and how they wanted, as opposed to GitHub where the only thing you can control is runners.
https://youtube.com/watch?v=tLdRBsuvVKc
It’s a shame they are selling out, but I suppose it was bound to happen sooner or later. GH is vacuuming up all of the market share.
I kind of liked the ability to self host my own GL. But for small/personal use cases. It was a bit overkill.
Like did calling it a “merge request” trick everyone into believing they weren’t a GitHub clone in the early days?
"pull request" is a more recognizable term, but "merge request" is more technically accurate for what GitHub and GitLab do.
And is it really jumping through hoops to just name it something different? I usually think of that phrase as implying something more arduous.
To that point, GitHub still does not use git-request-pull or even pulls to implement PRs, so the name is still not correct there. GitLab's naming is more accurate.
"Merge request" is a much better name for what GitHub actually does.
Imagine mathematicians having ten different names for "a curve", half of them considered offensive because why not.
You’d be surprised to learn that mathematics has at least three different names for a function.
Gitlab is for when you want to self-host a godawful corporate "CI/CD pipeline".
You either grind and love the grind, and make peace with your slice of the industry (Automattic [$7.5B value], for example) or you sell out and up when you get bored or tired.
https://finance.yahoo.com/quote/GTLB/
What do you think they should have done better that would have helped them (their bottom line, growth, brand, sustainability, or some combination of these)?
If you’re not the top 1 (or 2) in what you do, your fighting for extremely small slices of market share.
And lot of people still use AWS code whatever.
There's probably not much money in just hosting code (especially considering that the market has been "commoditized to death" by ... GitLab itself, among others).
It's a bit worrying to think about the impact a purchase/merger might have on the product.
What the article doesn't make clear is how much of the company capital is public? If the founder owns 45% of voting shares and Google/Alphabet has 22%, are only 33% of the shares actually on the public markets?
For example my bill with no additional needed features has risen from 6€ per user per month or so to €29 in something like 2-3 years.
I’m sure any serious buyer would look through that and realise some major churn is about to happen as renewals hit. - especially as they will be wanting customer/rev growth.
edit: I guess they do https://careers.datadoghq.com/remote/
I don't dislike Github but prefer Bitbucket, I feel they're a bit underappreciated, in part due to the massive market share that Github has. This could help turn things around.
I'm a GitLab user, I can't use Datadog because our customer contracts forbid data leaving the system (PHI, BAA won't suffice). I would reconsider GitLab if they were acquired.
Edit: 15 remote engineering positions open out of 219 total on their career site
Then, about 5 years ago I started working in a company that went the GitLab way, specifically due to the CI/CD support they provided directly in the platform.
At first, I didn't like the change so much, but as the project kept going, and we invested more and more in CI/CD, I can't see any going back. There's just nothing in GitHub that would allow us this kind of control and flexibly and automation over the build/testing/release process.
I get it if for your own personal development you prefer GitHub (I still do), but for anything of a medium/big size that's willing to invest time and money into a fully automated release process, GitLab is the way. And I guess that's where the value of GitLab to investors lies, in its appeal to the enterprise sector.
They really fucked up their pricing. There is such a steep cliff to climb these days. Literally starts at $30/user/month. You can almost get GitHub enterprise and Copilot for that price.
But their open core model, fully remote organizational structure, and overall product quality is highly valuable. How many open source distros self-host GitLab? Debian, Arch, Alpine, postmarketOS, any I'm missing? Then GNOME, KDE, and free desktop have their own instances. How many companies self-host or use gitlab.com? There is real value and momentum there. Would be a shame to succumb to impatience and AI frenzy. I doubt datadog can steward the project in the same way.
Gitlab has always been massive bloat that takes way too many people to maintain, and way too many resources to deploy. Not to mention proprietary feature gating that is a major turn-off to community contributions.
I wonder how long the new owners will add AI, Copilot, 2FA and other crap to their site.
I went github -> back to RCS and anon ftp -> just now to gitlab
I guess it is back to RCS and anon ftp.
FWIW, I would expect IBM to grab gitlab if it is really up for sale.
Why would that be the conclusion? By all means self-host, but you can self-host git (including just bare git on a server cloned over ssh) and it'll suck less than going back to rcs.
I use GitHub because there is no compelling reason not to and that’s what we have at work. But GitHub is stagnating and I have the idea of them having even more market share. Their AI propaganda is also insane.
Also supports Hg.
https://docs.github.com/en/enterprise-server@3.14/admin/over...