One thing that I has been wondering is while there are relatively large amount on alternative operating systems (and some are quite awesome like Haiku), still most Linux distros seem to be repackagings of classic gnu/x/(gnome|kde) apps. I find it hard to believe that these days high level userspace innovation is so heavily bound to kernel. Why don't we have more diverse Linux distros instead of having diverse collection of complete operating systems?
Especially for hardware support it would be kinda nice if more projects would use a common kernel instead of writing one themselves, and I find it also hard to believe that its easier to write a new kernel (that actually works in a range of hardware) compared to modifying an existing kernel to fit the project.
Indeed.. it is a sad state of affairs when GoboLinux ( http://www.gobolinux.org/ ), though innovative, call themselves the alternative Linux distribution. They are one of the few unique Linux distributions, and all they do is redefine the filesystem hierarchy.
There are an abundance of 'niche' distributions (focusing on size, audio support, etc) but none that really attempt new things. Actually, I'm sure there are some more but I haven't heard of them.. please inform me.
But of course, why are we complaining instead of thinking up possibilities of our own?
It's impressive. It feels like BeOS, unlike Zeta which was had poor quality.
There's some things not like they should be. This alpha is missing vi and ed (and no 've' either). It has nano. It has gcc and python2.6 and a perl5 though. Python libraries seem to work.
In some ways it's further ahead than Be. For example - far better web browser than was ever available back in the day (even relative to web browsing standards of the time). Could use some polish, but still a remarkable achievement. Better email functionality, although parts could be implement a lot better.
Little things are not right - keyboard shortcuts and the like. But it's much better than I was expecting on this front. People have definitely been putting energy into quality. A lot of little things work like they should that I've tested.
I'm using it under vmware on a macbook. The 'tracker' interface still feels way better to use than anything else I've used (including current-era finder).
I was a die-hard user, and it's quite novel to think that - long after having given up on the dream of being able to run this as my primary workstation OS - I may be on the brink of being able to do just that. The key will be whether it can be installed natively with all drivers working effectively.
I believe they deliberately removed vim from the default install since they felt that it wasn't suitable for a desktop-oriented operating system. It can be installed using via optional packages ('installoptionalpackage' in console).
"It's impressive. It feels like BeOS, unlike Zeta which was had poor quality."
I guess I should have clarified - I've never used BeOS either, or well, even seen it I think, so more detail would be interesting, but from what you're telling me so far its not in a very usable state, but nonetheless I'm curious, might install it some weekend.
I've installed it in VirtualBox. Pretty slick UI (although it does take some getting used too). Reminds me a whole bunch of most Linux\Unix distros. It's multitasking abilities has been hyped up much, although it didn't seem that mind-blowing to me (that might have been because it was running in a VM). Reportedly the OS code and API is very clean, so unless you are planning on developing on the platform I don't see why you would choose it as a primary OS over Linux.
Since you're trying it, I'd like to try and sell you from this perspective: it takes simple ideas that would be at home in the unix philosophy, but applies them to a graphical interface.
Try navigating around the filesystem quickly by right clicking on the desktop (holding it) and then following the directories. If you want to open one, just put the mouse over it and lift.
Something that didn't get much press but which is unique and wonderful to use is the integration between the filesystem and the UI. Email messages are just files that sit in a directory, but the tracker exposes attributes to them, and you can have it display columns for 'From' and 'Subject'. Your inbox is just a folder that contains your mail messages. There's an instant message client that works along these lines too. Users are just files, and the IM functionality is a plugin engine to the tracker. This feels connected to the 'everything is a file' philosophy.
The OSX filesystem is technically capable of some of this stuff (written by the same guy, and later, so it's probably a lot better), but the finder isn't. So even though OSX is probably superior internally, in practice it isn't better to use.
There's a library called bethon that exposes the internals of the OS to python. I've used this with success in the past, not recently. It used to be the case that you could use a combination of the C++ BeBook dev but then piece stuff together from the python console.
Update: Got bethon working. It's a bit of a fiddle in this VM because there's limited disk space available, and sed isn't linked in /bin. But once you get it built the example at http://donn.drizzlehosting.com/Bethex.html works fine.
I loved the BeOS -- happily discarded Mac OS 8 when BeOS PR1 came out -- and I am so happy to see Heiku OS on the front page of Hacker News. To this day, the BeOS is still the OS which brought me the most fun on a day-to-day basis.
Especially for hardware support it would be kinda nice if more projects would use a common kernel instead of writing one themselves, and I find it also hard to believe that its easier to write a new kernel (that actually works in a range of hardware) compared to modifying an existing kernel to fit the project.