Considering to move from a selfhosted gitea to Github with flockingbird. Sourcecode, issues, wiki/documentation.
* Discoverability: We're just more visible on GH than on "some random website".
* Participation barrier: you probably have a GH account, but certainly no account yet on our instance.
* Defacto standard: external systems like CI, support etc all require github.
It feels a bit like giving in, but the benefits seem to outweigh the idealistic stance.
What do you think?
Sourcehut has a CI setup, which we are still evaluating: our gitea needs work still to get a drone.io CI running.
So, did another comparison: Were are all the common fediverse projects at?
Using https://github.com/BasixKOR/awesome-activitypub, turns out that the vast majority is on github.
Se we'll just follow along, for now.
"Choose your battles wisely", the proverb goes. Distributing Open Source software project hosting is not our current battle.
@flockingbird GitLab has also a CI/CD setup and you can connect a GitLab.com account on a self hosted GitLab
@danslerush Thanks for the pointer!
git.webschuur.com, where flockingbird is hosted, used to run on gitlab. But it was too much effort & costly to run (I say that as a Rails developer and -hoster - Gitlab is built on Rails).
Going so far that on average we needed 0.1 FTE (half a day a week) just to keep it running. That is fine for a project with several FTE to spare, but not whith our focus 100% on the product.
Hosted gitlab might work, though, but is that much better than "hosted github"?
You could consider having just a mirror on GitHub and maybe allow authentication through GitHub on your Gitea instance. On the other hand if you don't mind the dependence one can't deny that GitHub provides a creepy amount of features.
Mostly features from third party tools, though. Even Open Source tools and projects have a "connect to github" but won't work with issues, PRs, wiki or other features on a gitlab instance.
I'm afraid we've dug our own hole here, as community. And locked ourselves into Github.
Yet, for flockingbird, our focus is our product and much less (the idealism of) the underlying infrastructure.
Still. It feels like "giving up".
Wanting to use MS GitHub for that has something "tasty"
I only pointed the CI/CD functionalities but you can add a layer on Gitea, like Jenkins or others /> https://docs.gitea.io/en-us/ci-cd/
But basically, it's Git everwywhere, you can mirroring ... or waiting for @forgefed 😜
@danslerush I have and had drone.io running on git.webschuur.com but it became too costly to maintain. Jenkins same: the TCO is just too high.
And nearly all (open source) hosted CI requires either github or gitlab and fails to run with a selfhosted Gitea.
So much for "Git - distributed source control" 😔
@flockingbird Yes I know, unfortunately, there is a big job to be done on this point 😥 I may repeating myself but GitLab remains a "less worse" solution in your goals
@danslerush I still have all the ansible playbooks to install & maintain that gitlab instance.
It was just so terribly resource hungry and therefore expensive (requiring a large VPS) and unstable that we moved to gitea.
Another reason was that gitea is much cleaner wrt its "client", the HTML. Making the SEO far, far better.
I guess gitlab.com migh prove a good middleground. Will investigate further. Thanks!
@flockingbird I have both, but yes I largely prefer my Gitea instance ... you're welcome ! 🖖
@flockingbird after Microsoft bought github, many people dropped their account or refuse to access it if they have one. The people using sourcehut or codeberg would much more likely be in that set.
As noted in the other thread, if discoverability is the concern, then following the Linux model might make sense (with pointers/direction to the official non-gh instance).
@flockingbird also, the people looking for a LinkedIn alternative will align with those avoiding github for the same Microsoft reason.
Just wanted to second what has already been outlined thrice:
If your target audience is people wanting a libre alternative to linkedin then they probably won't like to learn it's hosted on github (I know because I fit that description 😬)
Also, email contributions mean no account needed, and that's cool 👍
@ck I like that idea, and I did not know about that feature.
Is that something admins have to toggle per repo? Because I've not encountered it yet.
E.g. to add an issue at https://codeberg.org/Freeyourgadget/Gadgetbridge/issues I need to register at codeberg before being able to participate.
Which is quite understandable from codebergs' POV.
Yet it is another hurdle when we want participation to be as "frictionless" as possible. On GH, almost all potential participants already have an account, "sadly".
Nope. Just send a patch to the public mailing list of a project, that's it. Send an E-Mail to the public issue list, report an issue. No account necessary.
What about mirroring your repo to github (like the linux kernel maintainers do it) and have a bot that generates git send-email patches that they can send to you?
Otherwise what about selfhosted Gitlab?
Indeed, mirroring solves the "discoverability" part of the problem. We had that set up. But it introduces new problems.
The "Participation" becomes more complex: where do people post, update and search for issues? Where to update the wiki? How&where to update a PR? How to authorize push acces, etc?
We'll still need a canonical, clearly marketed, "single source of truth".
Should that be github? Or should everyone be directed to the selfhosted version, away from Github?
its a fucking shame that github got sold to microsoft. i hate this god damn company !
@flockingbird Rather than asking whether "the benefits outweigh the idealistic stance" (a framing that, I'd argue, prescribes the answer) I would instead consider what problems you will have to solve (make a concrete list!) if you do the idealistic thing and whether you think they can be dealt with, given your resources.
@freebliss Thanks for the suggestion.
We did that. And the outcome was clear: go for Github.
But that makes me unconfortable and feels like "giving up on free, decentralised web". Yet such a "feeling" is hard to place in a concrete list.
Hence the request for some other viewpoints here. Sorry if the "framing" comes over as such: it was an honest request for input on a matter that is important to us, yet hard to quantify; if at all.
@flockingbird I see. Well if you seriously "did the math" then even if it feels bad you should go with github! :), because anything else will, according to your considerations, hinder your project so much that it's of no help to the cause anyway. Failing idealistically is failure. If I misread things and you wanted feedback on your bullet points, mine would be that I would consider Discoverability and Participation Barrier both moot: Any serious and viably attractive project should and will (I believe) succeed without the extra discoverability GH provides, if it does not the proposition or marketing is probably too weak (note that this is no judgement on your project, I don't know it (yet)). Likewise any serious and determined contributor - the ones a project really needs - will not be hindered by having to register somewhere (my feeling anyway) and one shot "drive-by" contributions are probably negligible in the big picture. One shot issues/bugreports might be a relevant factor though, I concede to that.
@flockingbird I didn't mention these two points earlier because I guess your (or anyone's) evaluation on this will probably be decided in the realm of CI requirements anyway, this is where the real prohibitive factors can occur :)
@flockingbird Cool beans, Drew builds rad stuff. :) Fingers crossed!
@freebliss Thanks for the insights. I agree with the points probably being moot.
But we can find out. It's not as if a choice is made Forever.
In that light, I guess choosing something that offers freedom and flexibility to move when time arrives, is probably more important.
Points in which Github ranks low.
(Though unfortunately, for example, there is an easy migration from GH to any gitea, but not gitea to gitea. My last bullet point in action)
@flockingbird you could ask @yarmo who is creating https://keyoxide.org/ from #codeberg where I also have an account and several orgs. He has set up an self-hosted #drone CI environment that looks quite cool (the whole project does).
You might consider mirroring to GH, where you have the topic tags for findability, ppl can clone, fork, download releases, etc.
I intend to have some fedi projects both on GH (code) and codeberg (architecture discuss, docs dev).
What I did is having my main repo on #codeberg and an empty repo with a link to main repo on GH. So it's discoverable on GH. There's just a link you have to follow and an account to make if you wish to contribute. So far, it's working: people passionate enough made accounts to contribute. Am I missing out on contribs? Most likely. But that was my "maths". No drive by issues, no incendiary threads.
I get to wake up knowing what I'm doing is the right thing and not give in. And I'm promoting alternatives. If I want a future without GH, I need to make that future possible. I can't expect the world to move beyond GH if I can't make the effort to.
Now the big one: will my project fail because of this? It just might. But if so, I will have learned an invaluable lesson about hypocrisy and double down on alternatives.
@flockingbird Meet in the middle and go with Gitlab - great tooling, easy participation. Discovery isn't as good but it still has network effects and there's a lot of overlap between GH users and GL users: word gets around.
@flockingbird Also, Github is a product of Microsoft, and so is LinkedIn. It would be a hazard to rely on your competitors for critical infrastructure..
@flockingbird I agree, GitHub may not be FLOSS itself, but it definitely does host a very large number of project that we all depend on. If you wanna attract contributions, GitHub is certainly the best way.
@kriive @flockingbird @sedcat For how long ? « Ugh, this has Microsoft's playbook written all over it. Introduce a certification, thus increasing the gap between developers who (had their employee) pay Microsoft and developers that didn't. Conflate a generic concept (Git in this case) with Microsoft's specific implementation (Github), muddying the difference in managers' lexicons. Attempt to set Github as a standard to reach in everyone's mind » /> https://news.ycombinator.com/item?id=24855606
@kriive « GitHub was a standard way long before Microsoft acquired it » Yes, I'm agree/aware of that but it belongs to MS now and we can't deny it. Things are not going to get any better and if you want to build something in the long term you have to have a minimum of foresight. But yes, like I said it's Git, we can mirror, fork, copy or migrate when we want 🤓 @flockingbird @sedcat
I think you are wrong about "not our current battle" part. It's an ongoing battle for every software developer who care about own and users' freedom.
You can fight by avoiding corporate traps and unnecessary centralization. The earlier the better.
@sedcat It was phrased unfortunate. Sorry.
What I meant, was "it is not our *primary* battle", not the most important one. Flockingbird wants to bring "professional social networking" to the fediverse. This is the *one* thing that matters to me/us.
We are not a project who wants to "change the way open source developers work", nor a project who aims to "decentralise software development".
Thats a battle for others to fight, now. Projects like @forgefed, @codeberg, #sr.ht, @gitea or gitlab.
Fosstodon is an English speaking Mastodon instance that is open to anyone who is interested in technology; particularly free & open source software.