Mac-optimized Linux Distro

tl;dr: Who'd be interested in a Linux distro with a familiar OSX look-and-feel, decent performance on PPC Macs, intuitive user experience (App bundles / software centre instead of package manager to install apps) and a bright future through the latest software?


I'm part way through porting the excellent Crunchbang Linux to the PowerPC platform (I'll hopefully have some progress on my blog in the coming days) and am toying with the idea of an OSX-esque Linux distro specifically optimized to get the last drops of performance out of these machines. I would like to gauge the interest in such a distro, because it is no small undertaking.

Now there will be people from the Linux camp gathering up their pitchforks and torches, but the truth is that there are a lot of PPC users out there still (I've had over 1,500 downloads of my Flash 11.5 hack, I'd be interested to see how many downloads the Mountain Lion theme has had), and not many of them have the patience nor time to learn Linux. I know Ubuntu exists for this purpose (among others), but let's be honest, it's severely bloated. While the Ubuntu derivatives are much better (Lubuntu, Xubuntu), the user experience for a non-technical person is not going to be fun - especially as they don't hold your hand all the way like Ubuntu does.

The problem is OSX works wonderfully, but it is long abandoned. There is a core group of developers who still do some brilliant work (TFF etc.), but that is it as far as app development goes. I'd be hard-pushed to find many popular apps which are still developed for PPC - it seems they've all dropped off these past couple of years. So while the OS still works fine (let's not forget security issues mind!), we are stuck with the prospect of next to no new apps. People like the eye-candy, the way it just works, and the quality apps.

Such a combination is rare in a Linux distro and is why I am suggesting such a thing in the first place. Taking the best from both worlds is what I envisage. A familiar OSX (Tiger?) look coupled with a carefully thought out selection of software built with PPC optimizations and an intuitive way for managing apps (I wrote an OSX-style app bundle implementation for another distro I was working on) and I think you have a really nice solid base for a future-proof OS for daily use.

I started a thread the other day saying what technologies were keeping our Macs alive and relevant - I don't think any of them were tied to the OS. We have hardware which is still perfectly relevant and useful in this day and age (graphics hardware, USB 2.0, wireless etc.) as well as a rich and open platform for apps on the web (HTML5). Taking advantage of all of that in such a distro would be an amazing bonus.


Thank you for staying with me thus far, it was a lot to read! I'd be interested to hear comments.

GermanyChris said:

Pear Linux is most likely what your thinking, there is another but it's even less popular.

@OP Please don't mess up crunchbang it's a great distro making it easier to use will in some ways ruin it.

Click to expand...

This is a separate project from Crunchbang. The Crunchbang port will be as faithful as possible.

rjcalifornia said:

Ok:

1. Great effort! I always wanted to know how to compile your own distro

2. Crunchbang looks awful. I'm sorry but that looks real awful and just like another Linux distro. Yeah yeah I can 'change' the look but that's not the point. They should at least put some effort in making things smoother and prettier.

3. Although it is the most powerful, robust and reliable OS for Servers, for home users just sucks. Remember the Walmart Linux powered laptop fiasco? Return rate was over 100% and such laptops had a better RAM/Processor than our beautiful powerpc macs. Linux just cannot be as user friendly as Windows/Mac/Solaris

4. Apps. Unless apps can be distributed as stand-alone installers, the Linux you are working on will be a success. Installing an application in Linux is a pain when you don't have prerequisites or something like that. I don't care about installing prerequisites, but sucks when you have to install them.

5. Apps again. Are you gonna provide apps in the distro? Are you going to compile for specific laptop, like ibook G4 compilation or powerbook G4? Are you going to make use of Altivec?

Click to expand...

Crunchbang is an acquired taste and was never meant to be a mainstream distro. I personally love the minimalism and use it on all of my machines - but each to their own!

The aim would be to provide a full set of replacements for OSX apps installed by default (Firefox instead of Safari, Mplayer instead of Quicktime etc.) which will hopefully provide similar functionality and improved performance compared to what we use now. I will hopefully dig out my app bundle implementation and write a tool to convert Linux packages to 'app bundles' for easy distribution.

The kernel and all apps would be compiled on a G4 with PPC-specific optimizations applied wherever possible. Unfortunately I lack both the time and the expertise to fork packages and implement Altivec (though many apps already have altivec support built in, it just needs to be configured when building).

The problem I see is that PPC OSX developers are a dying breed - Linux developers are not.

YellowDog Linux? Gentoo PPC? FreeBSD/PPC? GNU/Darwin? Darwin? Mac OS X (which is a Darwin distribution)?

Just build a custom theme for the window manager, no need to port something.

Click to expand...

These are all fine, but dump an OSX user in front of one and they'd be quickly lost. It's not as simple as bundling a custom theme - what about all the media files they're used to using? How about the effortless connectivity and interaction with other Apple devices on the network? etc. etc. The point is that a user would want a very insignificant workflow change going from OSX to Linux which in its current state just doesn't happen. All of these services need to be configured and ready to go out of the box.


So I knocked up a quick demo and the results are rather pleasing. The system idles at ~55MB RAM and 4% CPU. Opening up an instance of Firefox on the MacRumors page totals 120MB RAM and a slight increase in CPU usage. Streaming a video from YouTube gives 157MB RAM usage and maxes out the CPU.

There's really nothing special going on here yet, it's just a really fast and light WM with a couple of cheap and nasty themes applied. The nice part is that there's a full compositing manager running, automatic service discovery, global menus, file indexing and a couple of other little tweaks. I am quite impressed with the very early metrics as there should be plenty of room to grow.

What's the source on those graphs - they look useful!

It's not so much to do with performance, than to do with the software available. I would like to use my PowerBook G4 for the duration of my degree until (or perhaps, if!) I graduate. At the moment I simply can't because it is very hard to find scientific/engineering software for PPC Macs - a problem which is non-existent with Linux.

At the end of the day it's basically a matter of how many people are happy with their systems and intend to keep them as they are (I was until I couldn't find engineering software for it), versus how many people are struggling because of the lack of recent software.

There is never going to be a magic performance increase - it's simply about having a decent selection of modern software available in the near future.

The usability of OSX at the moment literally hinges on the development of browsers such as TFF and Camino etc. for many users.

Not all MacPorts work on PowerPC unfortunately - Debian has a package for each of the major architectures it supports (except for non-free). Many of the apps that I use also bring in some horrendous dependencies - something I'd rather avoid having to build each time I wanted to install something! How many OSX users know how to work a package manager, let alone use ports?

Perhaps the solution is to start bringing the most popular Linux software and bundling it into OSX apps.