Who is Going to Buy SkyOS? 118
An anonymous reader wonders: "With the huge amount of operating systems available (numerous free and non-free Linux distros, Windows, Mac OSX, BSD, etc) who would buy SkyOS? An OS that was once free will now become a commercial operating system with the release of version 5.0. Although 'Porting applications from POSIX operating systems is an easy task', applications will still have to be ported since SkyOS 'isn't based on any other operating system'. This leads me to wonder...is there something about this operating system that I'm missing? Has anyone out there tried SkyOS and why would anyone pay for SkyOS with all of the alternatives out there with tonnes of software easily available?"
Re:Why? (Score:5, Informative)
Re:Why? (Score:2, Informative)
Re:I don't think so - at least for now. (Score:3, Informative)
"...This is the Crystal icon pack created by the very talented Everaldo, used in SkyOS with his blessing. This icon set is also used by KDE for Linux, which is why the icon sets look similar."
Re:I don't think so - at least for now. (Score:5, Informative)
As another posted pointed out it was with the creators blessing, however even that wasn't needed.
The Crystal Icon Set is licensed under the LGPL, so basically, as long as the SkyOS team supply you with a copy of the LGPL license and a written offer of the "source" (e.g. original pngs) they can use them in a commercial application.
There is a common misconception with some people that (L)GPL=no commercial usage. If it's GPL you can still charge (however your clients can turn around and distribute your app for free, so you'll get further charging for support). If it's LGPL you can distribute the rest of your app as closed source, as long as you provide the LGPL license and provide the source of the LGPL'd component(s) in some way for at least three years (you're even allowed to charge a reasonable fee for providing it under the terms of the license).
Re:From a Sky OS Beta user... (Score:3, Informative)
If you want an idea of why the parent's question is valid, download a Linux kernel source and look at all the subdirectories for drivers. It isn't trivial to make an operating system that appears to work on anything you throw at it!
Re:BeOS was a superior O/S... (Score:3, Informative)
Anyway, my proposal is:
1) replace processes with components.
2) replace the scheduler with parallelizing operators.
3) replace the filesystem with collections.
What are components? components are computational units that are maybe composed from other components and that accept an input and produce an output. The difference with processes is that components can be combined at run-time as the result of computation.
What is a parallelizing operator? a parallelizing operator is an operator that defines a) splitting of computations into parallel tasks and b) rendezvous points.
What is a collection? a collection is a set of data. A datum may itself be a collection. The difference with filesystem is that a) data are typed, b) since data can be collections, they can have triggers that notify the environment about changes. An arbitrary number of components can be attached to triggers. Data in collections can be marked as 'persistent'. Indices can be attached on collections for fast data retrieval. Persistence jobs can be modelled as transactions.
What are be the advantages of this proposed model the current model? they are:
1) program creation becomes much more flexible. There are no barriers between co-operation of components like a separate memory space, unless it is really needed.
2) extending applications becomes a trivial task.
3) code is separated from data.
4) the data persistence problem is solved.
5) programs can have database facilities available without cost.
6) software is componentized. Replacing faulty parts does not require replacing whole programs.
7) reacting to data changes becomes trivial.
8) better security since security can be defined at component level.
I would really like to see a pioneer O/S that utilizes some of the above ideas (which exist for a long time) instead of a repetition of the same stuff.
Re:Why? (Score:4, Informative)
I totally agree, I absolutely love Cosmo for these reasons.
I am so tempted to go out and get a subscription.
A couple of nice ideas: (Score:3, Informative)
1) An automatic media playback/manipulation framework. Nothing new here (see DirectShow, gstreamer, Quicktime) but it's more transparent, easier to configure they way you want to, it's an OS-level service, and it comes with lots of filters and encoders/decoders out of the box.
2) SQL metabase for your files. Very similar to WinFS or beagle/inotify in style. You write plugins to extract metadata and it indexes it when you make FS changes. And standard widgets to search/query for file dialogs, file browser, etc. Which is important...
3) A decent security context system that seems to be a blend between NT's model and SELinux's model. It uses the ideas of contexts for users, files and processes from SELinux, but it doesn't burden you with complex transformation rules (instead you have trusted right sources (typically the logon manager), inheritance and voluntary dropping of rights). It offers far more numerous rights than NT or SELinux, and you can add your own.
Files and directories can carry security context identifiers, and the system matches them from the least to most specific user identifier that matches for allow or deny (all users, group member, specific user). The default for filesystems that support security contexts is deny, for those that don't (IE VFAT on flash drives), the default is allow.
Again, you can find examples of all of this implemented elsewhere. These guys are trying to take a lot of good ideas and put them all together in a single baseline that is relatively easy to operate and get your head around. Which I have to support.
Would I buy it? Probably not. Not enough hardware support to bother with a new OS. Better to add these features to existing ones.