The distribution model is changing

One of the nice things about being on sabbatical is I get to say things like "You should be happy that RHEL is dropping LibreOffice RPMs" without the fear of being kicked out a room you didn't want to be in the first place.  

I have been talking for a while now about the economic benefits of aligning our desktop distribution models with that of cloud. After all, if a certain model is being way more successful than the others then the hard part is figuring out how to leverage that for the desktop.

If you do the napkin math, it's clear why Canonical moved to the snap model for Firefox. It's too expensive. There's no reason to have an entire body to do nothing but bring a browser to your OS. This is a really hard job (sup Chris Coulson!), because you're not just doing it for the current release, you're the poor soul that has to do all that backporting stuff for older releases.

What is easier long term for operating system companies is to build the frameworks for developers to do the self serve model. I would say "just like cloud" here but you already knew I was going to say that. Why bother doing all this stuff when you can just help upstream check in a yaml file and let CI/CD do the work?

Well that's just lazy

Yes, yes it is.

The idea that "correct" engineering is to keep the same inefficient processes sounds good on paper, but don't align with the economic realities. This classic distribution model isn't working for client, and I don't want it to work. Here's why I don't want it to work:

"focusing on gaps in Wayland, building out HDR support, building out what’s needed for color-sensitive work, and a host of other refinements required by Workstation users."

I don't know about you, but if a group of people have the skill to implement HDR and the proper color support that we know is barrier to professionals using Linux, and they were ... maintaining LibreOffice RPMs? This entire time?

Holy shit, I wish I could build a time machine to make them drop this two years ago. And while we're at it, we might as well just have anyone involved in doing anything in the critical path to just get rid of all that stuff from their lives.

The Value Proposition

Here's how the math works out: The value of "distributions" providing graphical packages for end users in client Linux is dropping dramatically. Those of us who have been using Flathub as their primary means of app consumption know this and have known this for years.

Chances your distro is better than Georges at hooking me up with that sweet OBS setup? Come on, why would you even want to do that when we can do it all together.

This isn't about us vs. them

There's an unfortunate misconception that this model is disrespectful towards distribution maintainers and that we should appreciate the hard work that they do and that it's a bad thing to remove this responsibility from them.

I 100% respect the work distribution maintainers do, I know hundreds of them. That's why they have the most important task of all. They're the only ones I trust to talk to the metal. And in order to do that we need to get all the other crap off of their plate. You know, those huge lists of orphaned packages that everyone pretends don't exist?

We need to move to a place where we recognize that in a world of limited resources that Linux distributions need to get past boiling oceans of packaging the world. Do what you excel at, providing us with a core desktop operating system with solid foundations.

People don't want to hear this so I'll just say it, distributions can no longer afford to pretend that they add value by repackaging office suites. The real value is in the hard stuff. We need a well maintained kernel, graphical stack, desktop, and associated core tools.

Everything else, punt to Flathub. Need something you can't get? You're in luck, every distro worth something also has containers you can use to get what you need.

But what about LibreOffice?

I think this is great for them. They get to publish to one spot, and that version is the same across all the distros. And when I say "all the distros" I don't just mean Fedora, Ubuntu, openSUSE, or whatever, we also mean across versions of distros.

All distros still need to care about LibreOffice, they're just not doing it in distro land. The ones doing it right will engage with upstream, that part doesn't change. What we're not doing is supporting a bunch of different versions across the entire universe.

Does moving to this model surface new issues? Yes. Are those challenging? Yes. Do we need to change how we do business around these packages? Yep.

It doesn't matter

At the end of the day it doesn't matter. The classic model is 5 people from 5 different distributions working on a package in their individual distros. The model moves it to 5 people working on common packaging and not having to deal with all of that toil so they can do more important stuff.

Is that how it will work in practice though? Not sure, but we know that the traditional model isn't working for client, so we don't lose much by trying.

Say what you want about Canonical and Red Hat's decision to do this, but at the end of the day it's all about the number of contributors you have in one hand, and the amount of work required to compete in the market in the other.

Expect this to happen more often, and for the pattern to accelerate because ... oh damn, you got me, this is what happened in cloud.

Show Comments