[kyran_adept]

I think Linus is at fault here. Having a PR process that requires sending a patch over email, email threads done with super old mailing list software, etc instead of the more modern workflow of using forks, PRs in a web UI, etc is tedious and annoying. Making kernel development and maintaining more accesible to younger people would increase the pool of potential maintainers.

[ssivark]

Keep in mind that linux is a 25+ year old project with established practices which might well outlast fads of web-driven development which might change every few years. Yes, Github/etc provide a friendlier interface for newbies (and might well be worth considering seriously) but note that there is almost no open source project that is on the same scale of collaboration complexity as the linux kernel, and it is far from obvious that they would be well served by popular platforms. Not to mention the whole Bitkeeper SNAFU that came out of becoming dependent on a platform they did not control; they would be very reluctant to put themselves in such a vulnerable position again. Github can decide on a whim to de-platform certain projects/customers/users (and has demonstrated a willingness to do so).

[Barrin92]

There's no need to rely on a third party provider given that it's possible to host say, Gitlab yourself. For a project of the size of the linux kernel that's more than a worthwhile investment.

[ssivark]

In this context it’s wise to move conservatively. Gitlab has “come of age” only in the last few years. We see that KDE is now moving. If that works well, I wouldn’t be surprised if other OSS projects slowly start migrating.

Remember that there is heavy organizational inertia and culture on the LKML. Moving to Gitlab is practically akin to gutting that and beginning a fresh project with forked code — the kind of cultural makeover that most companies don’t survive, and won’t do unless they’re forced to. And maintaining starting culture is even more crucial for volunteer-driven projects than for corporate projects with very different incentives.

[franga2000]

I really don't see a reason to not use both. Even right now, a maintainer of some subsystem could conceivably create a repo on Gitlab* and accept PRs there, which they could then (rebase-?)merge in and sync with git.kernel.org with Linus never having to know about it.

The problem is cultural, as well as organizational. Firstly, I doubt any of the current maintainers would be willing to do that, especially without Linus's blessing and secondly, without a fully managed Gitlab* instance that takes care of sync and whatever else automatically (provided by the Foundation), this would be an unreasonable burden on the maintainers.

[RMPR]

> I really don't see a reason to not use both. Even right now, a maintainer of some subsystem could conceivably create a repo on Gitlab* and accept PRs there, which they could then (rebase-?)merge in and sync with git.kernel.org with Linus never having to know about it.

FWIW that's already happening with the BPF subsystem https://github.com/libbpf/libbpf

[ta17711771]

Or Gitea!

[vlovich123]

I'd put LLVM/Clang, Chrome, and AOSP into those categories. They all have more modern development tools than e-mailing patch-sets around.

[finnthehuman]

I'm often confused as to how many people really have a potential future where they become motivated long term participants, but a few bumpy roads at the onset turn them off. And I don't just mean in kernel development, learning the guitar isn't fun either.

Who gets into software as a profession, and then lets a bad UI turn them away from a project? When so much of the valuable skillset in programming is understanding and navigating a possibility space where the UI isn't built yet? (And I'm sure some kernel contributors would argue it's more _unusual_ than actively _worse_).

I know this paints me as old and out of touch. I know this. But I think it still applies. When I look at my interns, the good developers tend not to be the ones tripped up by onboarding to systems and workflows that aren't exactly like the trendy development tools or what they saw in school.

[qchris]

I don't think this is an entirely fair comparison. I thinks it's closer to the Post Office being worried about finding enough new drivers for their mail routes, but refusing to upgrade any of their delivery trucks from a manual transmission to an automatic one.

Sure, you can argue that learning to drive a standard isn't that hard, and if you want to do a valuable public service like delivering mail, then learning to drive one is a small price to pay. But it's a barrier to entry for most young drivers because automatic transmissions are far, far more common. Besides, not only is the constant starting/stopping like a mail truck does usually hardest part of driving a manual, but isn't even fundamentally coupled to the act of delivering mail.

There's plenty of competent drivers who might be willing to deliver mail using an automatic (or developers to maintain the kernel on a platform with a decent UX), and that kind of process overhead isn't something that should just be waved away.

Edit: Changed accidentally repeated phrasing in middle paragraph

[pjmlp]

> automatic transmissions are far, far more common

The Post example doesn't work in Europe or Africa post.

[kstenerud]

If there were a better way to learn guitar, everyone would jump on it.

Software doesn't have to be so difficult and tedious to use, as has been demonstrated by a decade of u/x in the UI space. Nowadays, nobody would ever consider gordian knot for their video editing, and I don't blame them.

This idea that one must suffer to learn is as outmoded as alchemy. You don't need to take the good with the bad; learn from the bad and make more of it good. I lived though the old times, and they sucked!

Nowadays, if the developer doesn't give any thought to u/x, people won't give any thought to his creation, and rightly so. The days of cryptic incantations and tedious rituals are largely over.

[u801e]

> If there were a better way to learn guitar, everyone would jump on it.

This type of assertion is made in a lot of contexts, but I don't believe it holds true. I believe that there are a subset of people who are willing to learn a new skill. Of those people, there are very few who are willing to learn a new skill but won't because of some perceived barrier to learning it. The vast majority of that subset will learn the new skill despite the perceived barrier.

Of the people who aren't in the set of people who are willing to learn a new skill, there will be very few people who will be movitated to learn the new skill because the perceived barrier to learning it has changed in some way.

[imtringued]

I agree. Bad UI is not a barrier. After spending 3 weeks with FreeCAD I got used to all the weirdness. The hardest problem was finding good teaching resources and actually going over them. All you really need is a clear path for people who want to become kernel maintainers. If you rely on random luck it's obviously going to be difficult and nowadays there are more closed platforms than ever so less people even get to throw their dice.

[u801e]

> Making kernel development and maintaining more accesible to younger people would increase the pool of potential maintainers.

Is there an established open source project out there that originally used a mailing list or some other review software like gerrit, reviewboard, or phabricator, that transitioned to Github or Gitlab and actually increased the number of contributors and maintainers?

[TomJansen]

I could be wrong but ReactOS?

[kevsim]

I suspect he also deserves some of the “blame” but maybe not for the reasons you list. He has not exactly made the Linux kernel community a warm and welcoming place historically. Many people can probably think of more fun things to do than get publicly brow beaten by Linus

[ThrowawayR2]

Surely you jest. Compared to digging around in the kernel and device drivers, the current PR process is probably the most user-friendly and simple-to-use thing that kernel developers deal with in their work.

[pklo]

Of course Linus wrote git, so it's not like he's against modern workflow. There are kernel forks of course, I think that not using web front end is partly due to reluctance to become dependent on an commercial service that Linux people don't control, and partly due to the fact that they like to keep their hands in the mud.

[Havoc]

Flip side is there is something to be said for the Kernel not jumping on whatever platform is hot this month.

Plus I suspect the crowd capable of contributing meaningfully is quite comfortable with low level old school tech.

[effie]

Modern does not mean better. Email-based collaboration is most welcoming to new people and good enough for the regular participants. I for one do not want to create an account and learn how some slow web application works for each project when I can do with sending an email.

[icelancer]

>> Email-based collaboration is most welcoming to new people

What? Aren't new collaborators way more likely to be younger and acquainted with Web-based tools like GitHub/GitLab?

[effie]

New wannabe collaborators maybe. But does the project want to attract only young people who don't care about corporate control creep over free software? Or does it want to attract all able young people including those who understand the benefits of the email based collaboration, which is being the only convenient, decentralized, non-corporate-controlled and the most equitable way for strangers to communicate over the Internet?