Ask HN: What should an ideal, all-new OS look like?

3 points by Nomentatus ↗ HN
What would an all-new, best-we-can-do-today OS look like, if we hit restart and didn't have to worry about backwards compatibility to any significant degree? (Yes, I agree that's an extreme hypothetical!)

Would it be all THAT different?

In other words, would such a "best of show" OS (one that's within our technical reach, right now) look much like the incumbents? Would it be more serviceable, do more, or just be somewhat faster, smaller and more reliable? How much easier could it make life for programmers, if at all? Security? Future-proofing? How much economic value could doing such a thing actually add even if you could ignore legacy? (I know, not enough to actually dump legacy, but how much? Enough to make you pay 20% more for a computer, if that were the price you had to pay? 10%?)

Speaking of future-proofing, how likely are AI advances (deep learning) to heavily profit from such a rethink?

I'm hoping to avoid any discussion of why companies haven't done this and almost certainly won't: because I think we all know why such a project might not be profitable or very practical. I still use Qwerty, after all.

I love Rust, but since existing OSes can be rewritten in Rust and that may very well happen, if only a bit at a time; I don't see Rust as a decider, here. Of course, we want Rust, we just don't have to restart to get it.

This query was sparked a bit by this quote from romaniv: "One would think that in 2017 we would be using an OS that takes the best ideas from Xerox Park/Star, Oberon and Cloud 9. At least on new devices (like tablets) which don't have to keep backward compatibility." https://news.ycombinator.com/item?id=15704806

5 comments

[ 3.2 ms ] story [ 25.8 ms ] thread
Re Fuchsia: Google is restarting in one sense, replacing the Linux kernel in Android it would seem, partly in Rust. But they can't really restart, since they can't stray too far from their libraries and huge app pile. Is that something that eats at those programmers, or is it really not a big deal to forgo pure freedom to invent, since we've gotten it mostly right by now, anyway?

My strong impression of Redox was that it sticks very close to Unix principles (and libraries) and was not a clean restart. But I may very well not understand all that one can do if you are writing in Rust from the start.

There's a group of people writing a full, all-new OS from scratch in Rust called Redox[0] and Google is also working on a new mobile OS called Fuchsia[1].

[0]: https://github.com/redox-os/redox [1]: https://github.com/fuchsia-mirror

My apologies, matteing! I posted then heavily edited the above OP query, then added my comment on Fuchsia - before looking for comments and reading yours. Plus now I've edited the Fuchsia comment, too. So your reply may seem to others to ignore what I wrote - but that's not true, you were replying to a much simpler query. It was I who ignored (missed) your reply while busily typing. Next time I'll polish first, then publish (I hope.)
(comment deleted)
When I last looked into ReDox, it looked like one guy was making most of the changes (but that was in the summer), my kudos to him.