Language choice discussion

[lambdanil]

Hi everyone,

With AshOS further maturing I think it's a good time to reconsider our language choice.

When the project originally started I chose Python as the implementation language - simply because it's good for quick prototypes and has a lot of features built-in. But now the scope of the project is much bigger than first expected, so it's a good time to ask if Python is the right language for this project.

One of the disadvantages of Python is it's reliance on a language runtime - if the Python installation is somehow messed up, the entire system falls apart. For that reason a compiled language might be better suited.

Here's a small list of the pros and cons of different approaches:

Approach one - compiled language (C, C++ or similar)

Cons

• Means an entire rewrite, obviously very big changes
• We would have to distribute the precompiled binaries (we could however just host them as releases on the github repo)

Pros

• More reliable - not reliant on a language runtime = fewer points of failure
• Harder for the user to break
• Less dependencies - smaller base installation
• We could achieve better interoperability with the system, using native system calls, native library for interacting with btrfs, etc.

Approach two - shell scripts

Cons

• Shell scripts aren't very well suited for bigger projects

Pros

• It's more unix-y
• More modular

Approach three - keep using python

Cons

• Reliant on a language runtime
• More dependencies

Pros

• Much less effort than moving to an entirely different language
• Python is quick and easy to write, which makes it easy to contribute and add new features

---

I'd like to discuss the merits of these approaches, and choose whichever one works best for the project. Personally I think approaches 1 and 3 are the more interesting ones.

I'll be waiting for any input from our community!

[i2]

In this discussion, I would make a distinction between AshOS installer and Ash itself (refer to this discussion).

Installer can stay in Python as performance isn't important and python give us pluggability and ease of development.
Just to clarify for participants, this question is more relevant to Ash itself.

For Ash itself, I'd like to envision AshOS in C, and that's proper C not OOP C++.

[xplshn]

I would suggest Go, because it is (like)C but convenient. And it doesn't have the dependency nightmare of C, and it can be easily compiled as a static binary too, and there is also the fact that one can compile it using LLVM/GCC if using CGO, also Go is very readable, unlike C++.

[i2]

@ochibani In your experience, what has been the pain point(s) with converting ashos to Rust so far?

[ochibani]

It was replacing 'python-anytree' and I didn't fix this, as the code still depends on this Python library.😅

[i2]

I can see this, but one aspect that I'd like to be considered: Does rust run on all the OSes and architectures that Python supports?

[ochibani]

Why not?

[xplshn]

Another cons against Rust would be that it is quite contrary to this:

Python is quick and easy to write, which makes it easy to contribute and add new features

[xplshn]

We would have to distribute the precompiled binaries (we could however just host them as releases on the github repo)

Is there a serious problem with this? Statically compiling Rust is not a big problem, however, Rust binaries that are compiled this way are usually bigger than other language's because they bundle everything, not just the necessary symbols. C, C++, Go, Zig, etc, simply bundle the symbols needed, not everything.