Re: is there a version of unix written in Ada

[Shark8 <onewingedshark@gmail.com>]

On Wednesday, September 23, 2020 at 11:39:11 AM UTC-6, olivier wrote:
> > Thanks everyone, it is more than clear Ada can do an OS, from the 
> > realtime OS links.
> There has been some interesting development lately: 
> 
> - https://blog.adacore.com/cubit-a-general-purpose-operating-system-in-spark-ada 
> - https://github.com/RavSS/HAVK 
> - https://github.com/ajxs/cxos 
> - https://archive.fosdem.org/2020/schedule/event/ada_spunky/ 
> 
> Also, you can find a 'probably' complete list here: https://github.com/ohenley/awesome-ada#OS-and-Kernels 
> 
> Personally, I think: 
> 
> a) efforts should be organized around a single project, 
I have a lot of ideas about an operating system, as I've been toying with the idea for two decades, even starting to do an OS in Borland Pascal 7, which made things really nice as the executable was runnable both as a DOS program and booting to the bare-metal. That said, POSIX is a terrible idea, it mandates many things that [IMO] severely constrain the architecture, to include directory layout IIRC, and making another *nix is something that holds zero appeal to me; instead, I would advise having a very solid foundational framework and some "native" constructs that you're not going to find elsewhere.

> b) we should find some financing for such a project and 
That's a tall order.

> c) a lightweight, but formal design procedure should be put in place. 
True.

> Else, we will live with non-lasting attempts and a C base linux, forever unstable with new holes every 3 months, and some Rust that will inevitably makes its way in. 
And there you start hitting on one big issue: C is so ingrained in *nix that "making a Unix in Ada" is begging to fail. There are many Unix-ims that are simply bad design, like the ad-hoc text-processing imposed by the design of piping programs together. (Yes, it does result in security flaws.) — Instead we ought to work "inside out" compared to the Unix mindset: create the foundational types and environment, complete with round-trip text-serialization, and then use that serialization in the command-line... rather than forcing new programs to conform to ad-hoc outputs we can have a system that offers a standard (and type-aware) set of streams for input and output.

It might be wise to break out the Orange Book, too: https://csrc.nist.gov/csrc/media/publications/conference-paper/1998/10/08/proceedings-of-the-21st-nissc-1998/documents/early-cs-papers/dod85.pdf

> I do not have the pocket to finance such a project but I could definitely project myself as thriving to organize and bridge different developers. 
I could do some design write-ups, but they definitely aren't Unix-y.

> If by any chance you know of a billionaire that has better ideas than trying to go to Mars and/or build a bigger yacht than its neighbor... give him my contact. Thx.
Yep, there's always the financial aspect.