Re^2: Anyone interested in coding a _real_ Perl OS?

[spx2]

Either use JVM or Parrot. It sounds like an achievable goal to me. It all depends on how much manpower you have, how much time you have on your hands. Imagine that if you engage in such a project it could and most likely will eat all of your time, so some funding must come from somewhere, and if you got all of that covered you're good to go :)

However, there must be milestones.

Like

• first writing a parser for Perl, using PPI for example
• generating the "machine code" that fits into JVM/Parrot, then writing a huge amount of test
• writing basic utilities for manipulating files
• writing XS stuff that would interface with hardware or low-level stuff

I am sure Perl is not fit for writing an OS, except you would invest heavily in a highly-optimized parser for it and a compiler which would enable you to produce binaries.

But I would look at implementations in other languages first, BrowserUk has mentioned JNode. I will mention House which is written in Haskell.

Also check out Cleese which is a "an operating system written almost entirely in Python".

Some OS implemented in LISP here called LOSAK, or another one called Movitz

I'm sure you can find a lot more examples, I'd read some of the progress these guys had first to know what obstacles one would encounter in building such an OS.

But seriously this is a huge project, and needs a lot of manpower, it may have some kind of motivation such as porting all the web-based modules from CPAN and then using this brand new Perl-OS as a platform to run webapps from, that would be interesting, that's an achievable, but far-away goal.

[BrowserUk]

Either use JVM or Parrot.

I think Parrot would be a very poor choice relative to the JVM.

The latter has had millions of dollars spent on it and it is very, very advanced. And they just turned out a release candidate for Java 7. It is proven technology that runs pretty much everywhere. And dozens of non-Java languages have been implemented on it. Many of those are in the public domain and so are ready examples of how to go about it.

My point about JNode is that they have done all the hard work. If you write a Perl interpreter that targets the JVM, the rest of the job is already done by JNode.