“C# has been a language with a mixed history but precise goals. Although the C# language definition is for some time an ISO standard, only a part of the Base Class Library, which contains the fundamental functions that are used by all C# programs (IO, User Interface, Web services, etc) is also standardized. Parts of the BCL have been patented by Microsoft, but that has not deterred developers from attempts at implementing the components that are standardized, in various forms (Mono and affiliated projects). What happens when you go beyond that? What happens when outside the language, you start to implement not a mere application platform, but an entire operating system around it? Brace yourselves, because there is not only Microsoft Research who has done this with Singularity, but at least two other projects doing the same; and they are doing this under opensource terms. A system based around a C# Kernel. In this article, we are looking at one of the two, Cosmos by asking Scott Balmos and Chad Hower about the project they are involved in.”
* Modern language (C#)
* Modern, professional license (BSD)
* Using Microsoft’s open-source programs, resources, and specifications to develop an OS (oh the irony!)
* Actual development team (and not a one-man ego trip!)
* Awesome design metaphor (C# to create an ‘Operating system Lego set’?)
* Running on real hardware (thus one of the first C# OSs to get beyond the theoretical stage)
* Complete CIL support — and because of this, you can (theoretically) write OSs in Basic, Python, et multiple al, and use this tool to develop it. (Edit: PHP? A PHP OS? @_@)
I didn’t even know about this one… and it’s something we can only hope goes further. ^_^
[i]Edited 2008-01-31 00:49 UTC
Hrm there is also SharpOS (of which I am a developer) which has been running on real hardware for a long time now.
What a fantastic solution for embedded systems, assuming they keep resource usage under control (which I’m sure they are).
What is the problem with C++ or C and why use a Microsoft-only framework/language (yes, Mono is not a complete and viable alternative for non-windows operating systems) ?
Because they can? It might prove useful, fun or just educational. I certainly like doing some things just because I can.
The problem is that C and C++ are obsolete languages that are a pain to work with. Thats like saying why use python when we have a perfectly good fortran?
And mono is a complete and viable solution for non-windows operating systems. If you don’t think it is, I seriously doubt you have ever used it.
Trolling ?
How are c/c++ obsolete languages?
‘c’ is arguably still the best “high level assembly” language out there.
exactly, an high level assembly it is. people are switching to managed languages because the cpu power/memory is there and it shortens development time a lot. it’s obsolete because for the really low level stuff you use assembler and for the high level stuff c#/java or something like that.
the performance gain/longer development time you get from using c compared to c# (or comparable) is nearly never justifiable.
the only reason c is around is because of legacy code or stubborn coders, not because it’s the best language for the job.
(yes, i’m generalizing, please don’t start showing me the one exception you can think of)
Those are a few of the appealing reasons, but just because code is managed doesn’t mean you have to take a noticable performance hit. There’s other appealing parts like builtin garbage collection, interfaces, code introspection, etc etc.
I am not trolling. C was a great language in its time, C++ was a hack that kinda brought it up to date, but both are very niche languages nowadays. What managed languages give in stability and security more then makes up for the relatively minor hit taken in performance, especially on todays systems.
It is the same shift that happened years ago from ASM to C. Eventually machines hit a point where it is almost universally a better choice to have the compiler take care of certain kinds of things for you.
It depends. “The relatively minor hit taken in performance” is highly subjective and, in some cases, unacceptable for some applications. But it depends entirely on the scenario and, in most cases, managed languages are adequate.
Actually, projects like SharpOS and Cosmos are aiming to use software-isolated processes to *increase* performance, especially the performance of context switching between kernels, drivers, and applications. And from all of the performance testing I’ve seen, C# is roughly equivalent in speed to C++ code.
Despite the comment threads already going on, there are a few reasons why C/C++ isn’t being used:
1) To prove it’s possible to write a CLR-based OS
2) The fact that an OS based around a memory-safe language and runtime is inherently more safe. You eliminate the majority of normal bug openings (buffer overflows especially). And without needing to use segments and separate CPU privilege levels, you don’t have to do context switches.
3) Most interesting to me is that by distributing code in a bytecode form, you get platform independence without having to distribute source. Device manufacturers can write a driver, distribute in bytecode, and not have to compile for every architecture, kernel version, etc. Nor do they have to distribute source, which somewhat protects trade secret interests (I know, bytecode can be easily reverse-engineered. But it’s a “see no evil” effort, in my opinion).
Hope that made vague sense.
Makes perfect sense. Between Cosmos, SharpOS ( http://www.sharpos.org/ ), and JNode ( http://www.jnode.org/ ), I think we all have a bright future ahead of us.
We won’t even have to make the mistakes of eternal backwards compatibility that Windows makes, because we can efficiently emulate other OS’ ABIs using OOP abstractions and opcode recompiling.
All hail open-source. Hallowed be the architects.
First, C# isn’t a “Microsoft-only framework/language”. It’s a public ECMA standard. See mono-project.com.
Second, C# has a lot of draws over C/C++: lower complexity, platform-independence, garbage collection (no more leaked pointer references), consistent advanced language features (ie. threading, exceptions, events, timing, delegates, templates, etc), rapid prototyping, JIT compilation (and support for on-the-fly compilation of classes).
Third, I don’t think it’s a question of “Why use C# over C/C++?” but, rather, “Why use C# over Java?” Computer languages are just abstractions. Java and C# can be reduced to native code just as readily as C/C++, and they contain all of the necessary constructs to write an operating system. Java and C# are both up to the task. I’m not sure why the project chose C#; except, of course, that tool support for C# is very good with Visual Studio (I don’t care for Eclipse–just my personal preference).
I totally disagree. Not sure how you formed this conclusion, but you’re wrong. Mono is an excellent open source implementation of C#.
So what language is the runtime environment written in? Somewhere you have to get down to runtime-free native machine code. Is there anyway to put C# on the bare metal?
From the website:
Cosmos includes a compiler (IL2CPU, which is part of Cosmos) that reads the input file (usually the shell) and Cosmos libraries and compiles the resulting IL to x86 code.