Schaller: Preparing the ground for the Fedora Workstation

You don't like it, fine. Some people do. Leave each to their own.

I agree with your conclusion, but not for the same reasons.

> I'm also deeply sceptical of his opinions on webapps and the “cloud”. We know that the US and UK governments and others do an insane amount of spying. Before we think about desktop integration, we need to think about how to make sure that those webapps don't spy on us! We need end-to-end public key cryptography everywhere *before* we get involved with webapps.

Unfortunately, the webapps battle was over long before most of us even realised it had started.

At the most fundamental level, webapps are *services* provided by some third party. No matter how much you secure your local system/client, said third party is completely outside of your control. The only way to ensure your data is protected is to run those services yourself, on systems you control completely. Which costs money *on an ongoing basis*.

Because that would mean that Linux distributions would have to give up control.

The mentality were a distribution spends 10 months pre-packaging everything under the sun and releasing it as some great eruption of software updates every 10 months is a broken model.

It creates a huge swath of bugs. It eliminates the user's ability to install the software they want with the versions they need. Right now if you have some piece of software on your current Linux system that you need to have a updated version of the easiest and most common recommendation for solving this issue is to _reinstall your entire operating systems_ (or do a major upgrade, which essentially amounts to the same thing).

And what if, god forbid, there is a major regression in the software that is released as a new update and/or you need to run a older version of the software for whatever reason. There is no formal solution for this. As a user you are completely left in the dust.

And the 'rolling upgrade' type distribution, like Debian unstable, isn't any better. Always having to be forced to run the latest version of the software isn't any better solution then always being forced to use the version of software that happened to be the most convenient for the distribution to package 7 months ago.

I like Linux and I like Linux distributions, but the model for managing and updating software is ends up being a sort of 'soft tyranny' over users. Most of the time it's unintentional, of course. All of this is just stuff that happens. Nobody is evil or incompetent here.. it's just how things are.

If you want it to be easy to install any deb or rpm on any system then what is required is a Linux standard base AND to eliminate the traditional Linux approach that they need to try to pre-package everything under the sun. Instead it must be made easy for software authors to package and distribute their own software themselves and make it easy and supported that users can arbitrarily pick and choose the source of their own packages and use whatever versions of software they feel they need to use.

Linux distributions can, of course, continue to work with upstream authors to improve packaging and create package collections for users to easily find and install software.

I really don't see this happening though.

So the next best thing you can do is simply side step the whole distribution problem and use a new packaging system that is completely separate from the existing ones. Let the distributions continue to play in their own private ecosystem, while you use various techniques like sandboxing, containers, or abstractions to help isolate the users from the damage that distributions and library authors inflict on software compatibility. This way you create a universal software distribution method that is no longer tied to particular OS releases.

Hopefully then you can create something that is so useful and easy to use that end users stop trying to run 'apt-get' or 'yum' to fetch the software they need and start preferring your own 'universal' system. That way distribution authors themselves will begin so the value in creating a layered OS and take a huge burden off themselves and actually improve their own ability to change and improve the system.

At least debs tend to install on any deb-based distro. That certainly didn't used to be true for rpms ... although Red Hat rpms did tend to work on Red-Hat-derived distros.

The only reason debs work everywhere debs are used is that all those distros are Debian derivatives and stick close(ish) to their parent. The two major rpm-based distros are not related (and indeed, Red Hat is the *younger* of the two). SuSE is a Slackware-derivative, although it's evolved an awful lot since then. So although it shares the same packaging format as Red Hat, it contains a lot of non-RH heritage.

If Lennart can successfully impose something similar to the LSB, it will give a major boost to the linux ecosystem - not least because it will make software vendors lives easier (and no, Free Software does *not* work that well in a fair few niches - if customers want to pay then Free doesn't cut it :-(

Cheers,
Wol

Cheers,
Wol

You're also completely free to propose, design, and build your own system, building on as much or as little of the current status quo as you see fit.

Good luck.

Oh, one more thing:

> Any project where a RH employee is top dog needs to be avoided like the plague by the larger community...

I guess that means you won't be using gcc, binutils, glibc, or even the Linux kernel itself then. Try not to let the door hit you on your way out.

It's interesting that you specifically mention Red Hat, since Ubuntu has also been following this same strategy for several years. One might say that RH has to do this to prevent Canonical from taking complete control of Linux.