[seba_dos1]

It's not just that "people are hard" - it was clear that this will end up this way the moment marcan started ranting on social media about having to send kernel patches via e-mails. Collaborating on software development is a social activity and stuff like convincing maintainers to trust you and your approach is just as important part of it (if not more important) as writing code. Not realizing that is a sure road to burnout (and yes, I'm just as guilty of that myself).

[bayindirh]

> Not realizing that is a sure road to burnout (and yes, I'm just as guilty of that myself).

Humans are shaped by experience. This is both a boon and a curse. I have been also been on the hot end of the stick and burned myself down, sometimes rightly, sometimes wrongly. Understanding that I don't want to go through this anymore was the point I started to change.

> Collaborating on software development is a social activity and stuff like convincing maintainers to trust you and your approach is just as important part of it (if not more important) as writing code.

Writing the code is at most 5% of software development IME. This is what I always say to people I work with. I absolutely love writing code, but there are so many and more important activities around that, I can't just ignore them and churn out code.

[josephg]

> Writing the code is at most 5% of software development IME.

This really depends on what you work on. And how good the managers are on your team. I talked to a manager at Google once about how he saw his job. He said he saw his entire job as getting all of that stuff out of the way of his team. His job was to handle the BS so his team could spend their time getting work done.

This has been my experience in small projects and in very well run projects. And in immature projects - where bugs are cheap and there’s no code review. In places like that, I’m programming more like 60% of the time. I love that.

But Linux will never be like that ever again. Each line of committed code matters too much, to too many people. Is has to be hard to commit bad code to Linux. And that means you’ve gotta do a lot of talking to justify your code.

I did some work at the IETF a few years ago. It’s just the same there - specs that seem right on day 1 take years to become standards. Look at http2. But then, when that work is done, we have a standard.

As the old saying goes, if you want to go fast, go alone. If you want to go far, go together. Personally I like going fast. But I respect the hell out of people who work on projects like Linux and chrome. They let us go far.

[BeetleB]

Even in the Google example, it's still in the low percentages when you view it as a system. All the manager did was efficiently allocate resources. It didn't reduce the non-programming work - it simply moved it elsewhere.

[whstl]

Not quite.

Someone who is in a management position, has good political skills and good connections will be way more efficient at doing some of this non-programming work.

This is something that even C-levels forget. Something that takes a CTO 2 minutes to do can take several months for a regular developer to achieve, and I have plenty of experience on and plenty of examples of that.

[josephg]

Yeah. I think the whole drama around rust on Linux is a great example of this. If Linus came forward and clearly supported (or clearly rejected) rust on Linux, it would have saved a lot of people months of stress and heartache.

[whstl]

Exactly, that's a perfect example.

[prmoustache]

It really depends what kind of code and for which usage.

People might also live their hobby dev experience better if they were really coding for themselves without any expectation except pushing the code to a repo. As a hobby dev, you don't have to make package, you don't have to have an issue tracker, you don't have to accept external contributions, you don't have to support your users if you aren't willing to have this on your shoulder. You don't even need a public git repo, you could just put a release tarball when release is ready on your personal website.

[seba_dos1]

This works perfectly fine as long as you're happy with being approximately the only user of your code. With some of my projects I do just that, but it gets very different once you add users to the mix.

[rowanG077]

5%? Sure there is a lot of activity around software. But out of week of 40 hours I most certainly code more than at most 2 hours. If this is your workplace I think it's dysfunctional.

[bayindirh]

Not by time, but by weight. The code you write is worthless if you can’t communicate it or what you did.

[rowanG077]

You are implying that if you can communicate but have nothing backing it up that's worth 95%? If anything code can still be taken as is and understood by someone else. So to me it's always most important to be able to produce anything before being able to communicate.

[bayindirh]

In a sense, yes. I'm contributing to a small but crucial part of a big project, as a coordinator of a four person team. The dynamics of the project form the team as "band of equals", in other words, everybody has approximately the same capabilities, and roles are divided organically, and I ended up as the "project manager" for that group somehow.

Even before we started coding, there was an RFC written by us. We have talked about it, discussed it, ironed it out with the chief architects of the project. When everything made sense we started implementing it. Total coding hours is irrelevant, but it's small when compared all the planning and it's almost finished now.

The code needs to tap and fit into a specific place in the pipeline. Finding and communicating this place was crucial. The code is not. Because you can write the most sophisticated code in the most elegant way, but if you don't design and implement it to fit to the correct place, that code is toast, and the effort is a waste.

So yes, code might be the most enjoyable (and sometimes voluminous) part, but it's 5% of the job, by weight, at most.

[sophacles]

Once your proof of concept gains traction more time is spent in meetings with other teams responsible for the systems you'll be interacting with - making sure you do it "right" rather than scrappy. Once your initial release starts getting popular you spend more time looking at metrics and talking to operations folks to make scaling easier and not waste resources. Once your product starts having customers who depend on it, you spend a lot of time working with product to figure out features that make sense, advise on time estimates, mentor new team members, and advise teams who use your product/services as a dependency.

These are all engineering tasks, and the longer you spend on a team/in a company, the more likely it is you provide more value by doing this than by slinging code. You become a repository of institutional knowledge to dispense.

[coldtea]

>You are implying that if you can communicate but have nothing backing it up that's worth 95%?

If you can communicate it can be 99% of the value. Getting someone to write something to back it up is trivial in comparison.

[rowanG077]

    "Talk is cheap. Show me the code."
    - Linus Torvalds

[hnaccount_rng]

Think about it the other way around: How much code is written and never used? How much code is written and would be better if it were never used? How much code is used only to then notice, that it doesn't solve the business problem that it was intended to solve? How much code is run and it's never noticed that it doesn't solve any business problem?

All the while: You are correct, being able to produce anything that solves a problem is much more valuable than being able to talk about it. But in order to unlock the value (beyond solving your own problem) absolutely requires communication

[seba_dos1]

It's more like writing the code is just the first step on a long road. You won't go anywhere at all if you don't take it, but if that's the only thing you do, all you did is the first step.

I have written plenty of code that's stuck on this first step in my life, including some that went to the very same LKML we're talking about here right now. Some of those things have already been independently written again by other people who actually managed to go further than that.

[tuna74]

It is not useless if the code is being run and is doing something productive

[BeetleB]

It very often is.

Perhaps "useless" was the wrong word the GP used. "valued" may be better.

It's fairly common for very useful/valuable code to be discarded because the engineer (or his management) failed to articulate that value to senior leaders as well as someone else who had inferior code.

[adamc]

This. So very much this. If you burn bridges and then need them later, yeah, things are going to be hard.

[mschuster91]

> it was clear that this will end up this way the moment marcan started ranting on social media about having to send kernel patches via e-mails. Collaborating on software development is a social activity and stuff like convincing maintainers to trust you and your approach is just as important part of it (if not more important) as writing code.

Yeah but FFS using email for patches when there are so much better ways of doing development with git? The Linux Foundation could selfhost a fucking GitLab instance and even in the event of GitLab going down the route of enshittification or closed-source they could reasonably take over the maintenance of a fork.

I get that the Linux folks want to stay on email to gatekeep themselves from, let's be clear, utter morons who spam on any Github PR/issue they can find. But at the same time it makes finding new people to replace those who will literally die out in the next decade or two so much harder.

[xorcist]

It's an interesting phenomenon where people keep coming out of the woodwork and criticize the most successful software development project in history for doing it wrong.

They're not micro kernel! They're not TDD! They're not C++! They're not CVS! Not SVN! Not SCRUM! Not Gitlab!

Yet the project marches on, with a nebulous vision of doing a really useful kernel for everyone. Had they latched on any of the armchain expert criticism of how they're doing it wrong all these years we wouldn't be here.

[mschuster91]

> Yet the project marches on, with a nebulous vision of doing a really useful kernel for everyone.

The question is - how long will it march on? The lack of new developers for Linux has been a consistent topic for years now. Linus himself isn't getting younger, and the same goes for Greg KH, Ted Ts'o and other influential leads.

When the status quo scares off too many potential newcomers, eventually the project will either wither or too inexperienced people drive the project against a wall.

[CRConrad]

> Linus himself isn't getting younger

Why would he need to, he's already a young whippersnapper.

[seba_dos1]

> there are so much better ways

...which doesn't matter at all.

The people in charge decided on their preferred ways of communication. You may believe that there are better ways out there, and I may even agree with you, but ultimately it's completely irrelevant. People responsible decided that this is what works for them and, to be honest, they don't even owe you an explanation. You're being asked to collaborate in this specific way and if you're unable to do it, it's on you. If you want to change it, work your way to become a person who decides on this stuff in the project, or convince the people already responsible. Notice how neither of those are technical tasks and that they don't depend on technical superiority of your proposed methods either.

[WesolyKubeczek]

> Yeah but FFS using email for patches when there are so much better ways of doing development with git?

You are missing one point, namely that email is probably the only communication medium that's truly decentralized. I mean, on most email providers you can export your mailboxes and go to someone else. You can have a variety of email clients and ways to back up your mailboxes. No git clone, no specific mailbox or server is in any way special, I think Linus emphasized recently that they made efforts to ensure kernel.org itself is not special in any way.

Yes, I find Github's or Gitlab's UI, even with all enshittification by Microsoft and whatnot, better for doing code reviews than sight-reading patches in emails. And yet I cannot unsee a potential danger that choosing a service — any service! — to host kernel development would make it The Service, and make any migration way harder to do than what you have with email. Knowing life, I'd say pretty confidently that an outcome would be that there would be both mailing lists and The Service, both mandatory, with both sides grumbling about undue burdens.

Have you ever been in a project which had to migrate from, say, Atlassian's stack to Github, or from Github to Gitlab, or vice versa? Heck, from SourceForge + CVS/SVN to Github or similar? Those were usually grand endeavors for projects of medium size and up. Migrate all users, all issues, all PRs, all labels, test it all, and you still have to write code while it all is happening. Lots of back-and-forth about preserving some information which resists migration and deciding whether to just let it burn or spend time massaging it into a way the new system will accept it. Burnout pretty much guaranteed, even if everyone is cooperating and there is necessity.

But you could probably build tools on top of email to make your work more pleasant. The whippersnappers who like newer ways might like to run them.

[bayindirh]

I personally don't think GitHub's PR model is superior to e-mail based patch management for two reasons. First, e-mail needs no additional middleware at git level to process (I can get my mails and directly start working on my machine), plus e-mail is at least one of Git's native patch management mechanisms.

This is not about spam, server management or GitLab/Gitea/whatever issue. This is catering to most diverse work methods, and removing bottlenecks and failure points from the pipeline. GitLab is down, everybody is blocked. Your mail provider is failing? It'll be up in 5 minutes tops, or your disk is full probably, go handle it yourself.

So Occam's razor outlaws all the complex explanations for mail based patch management. The answer is concise in my head:

> Mailing list is a great archive, it's infinitely simpler and way more robust than a single server, and keeps things neatly decentralized, and as designed.

This is a wind we can't control, I for one, am not looking and kernel devs and say "What a bunch of laggard luddites. They still use e-mail for patch management". On the contrary, I applaud them for making this run for this many years, this smoothly. Also, is it something different what I'm used to? Great! I'll learn something new. It's always good to learn something new.

Because, at the end of the day, all complex systems evolve from much simpler ones, over time. The opposite is impossible.

[noname120]

> Your mail provider is failing? It'll be up in 5 minutes tops, or your disk is full probably, go handle it yourself.

Well until you deal with email deliverability issues, which are staggeringly widespread and random. Email were great to send quick patches between friends like you'd exchange a USB key for a group project. For a project the size of Linux? It doesn't scale at all. There is a reason why Google, Meta, Red Hat, and [insert any tech company here] doesn't collaborate by sending patches via email.

[lathiat]

Just wanted to back up this comment with a tiny bit of possible evidence: https://lwn.net/Articles/950567/

[black3r]

the problem with mail-based patch management is that it doesn't scale well, management wise... when you have hundreds of patches and multiple reviewers who can review them, Github/Gitlab environments make it easier to prioritize the patches, assign who will do the review, filter the patches based on tags, and keep track of what wasn't reviewed yet...,

mail-based patch management is fine for smaller projects, but Linux kernel is too big by now.. it sure is amazing how they seem to make it work despite their scale, but it's kinda obvious by now, that some patches can go unnoticed, unprioritized, unassigned, ...

and open source is all about getting as many developers as possible to contribute to the development. if I contribute something and wait months to get it reviewed, it will deter me from contributing anything more, and I don't care what's the reason behind it. the same goes for if I contribute something and receive an argument between two or more reviewers whether it's the right direction or not and there's no argumentative answer from a supervisor of the project and this situation goes on for months...

[seba_dos1]

> and open source is all about getting as many developers as possible to contribute to the development

[citation needed]

It's what "open source" enables, but it may not necessarily be a desired goal of a FLOSS project.

[xorcist]

Email is just the protocol. What you're really saying is that http-based protocols make more powerful tools possible.

It's not really enough to state your case. You have to do the work.

On the surface, the kernel developers are productive enough. Feel free to do shadow work for a maintainer and keep your patch stack in Gitlab. It it can be shown the be more effective, lots of maintainers are going to be interested. It's not like they all work the same way!

They just have a least common denominator which is store-and-forward patch transport in standard git email format.

[mschuster91]

> GitLab is down, everybody is blocked

Everyone still has at least the base branch they're working on and their working branch on their machine, that's the beauty of working with Git. Even if someone decides to pull a ragequit and perma-wipe the server, when all the developers push their branches, the work is restored. And issues can be backed up.

> Also, is it something different what I'm used to? Great! I'll learn something new.

The thing is, it's harder and more difficult in a time that better solutions exist. Routinely, kernel developers complain about being overworked and onboarding of new developers to be lacking... one part of the cause certainly is that the Linux kernel is a massive piece of technology, and another one that the social conventions of the Linux kernel are very difficult, but the tooling is also very important - Ballmer had a point with "developers developers developers".

People work with highly modern tools in their day jobs, and then they see the state of Linux kernel tooling, and they say "WTF I'm not putting up with that if I'm not getting paid for it".

Or to use a better comparison... everyone is driving on the highway in the same speed, but one car decides to slow down, so everyone else overtakes it. The perpetual difficulties of many open source projects to accomodate changing times and trends - partially because a lot of small FOSS is written by people for their individual usage! - are IMHO one of the reasons why there is so much chaos in the FOSS world and many private users rather go for the commercial option.

[adamc]

You are missing the entire point. When you interact with a group of people who already have a culture and a set of practices/traditions, you have to play by their rules, build up credibility with that community... and then maybe, down the road, you can nudge them a little to make changes. But you have to have credibility first, have established that you understand what they do and understand why their preferences are the way they are.

If you approach it from the viewpoint that you have the solution and they are Luddites, you will influence no one and have no effect.