It's certainly an impressive feat of engineering, but I'd really like to see a new OS project that isn't a Unix clone. OS development is a monumentally large endeavour and in my eyes it seems like a waste to dedicate that level of energy to copy something in a new programming language.
Yep, this is the technical debt OS creators are facing. There was actually a time not that long ago when operating systems deviated wildly and struck out in all sorts of directions, but because of lack of applications most of them died out despite the immense amount of work put in to them.
If you’re not building in some amount of compatibility, building the OS itself is a fraction of the work ahead.
why is it that any mention of Rust on the internet will bring about these wildly-overshooting claims. The amount of production software having been re-written in Rust (or replaced by a Rust counterpart) is minimal. This kind of rewrite is rare for any programming language. Yet claims like this are everywhere.
I find it annoying either way: it doesn't provide valuable/novel information and isn't conducive to productive conversation (in part because it's unclear if it's sarcastic or earnest)
Sort of. Is the Linux kernel huge and has an amazing number of man years in it .... certainly.
If starting over from scratch could you ignore most architectures, drivers, etc and support current hardware ... also yes.
A new OS could skip 32 bit mode, alpha, mips, itanium, loongarch, 68k, pa-risc, powerpc, s390, superH, sparc, parallel ports, serial ports, supporting zillions of PCI/PCIe cards that you can't buy any more, messing around with archaic 16 bit bootloaders, etc.
I think it is a Ship of Theseus situation. Once Rust becomes acceptable in the core kernel, module by module, it is going to spread.
I am fully Team Rust for core infrastructure, but I am curious how many security blunders in the kernel are memory errors vs bad security architecture.
If you're going to write a whole kernel in Rust, why would you settle for making a copy of Linux? Bug-for-bug compatibility with 30 years of cruft isn't exactly exciting or even necessary.
Reposts are fine after a year or so (this is in the FAQ: https://news.ycombinator.com/newsfaq.html) but the most recent thread was much less than a year ago, so the current post counts as a dupe.
Are you planning a "previous submission" bot which automatically adds a list as the above to all resubmissions? Would save you some time and give immediate insights to those (re)interested in rather frequently posted urls and topics.
More likely a way for the community to build these lists collaboratively. The trouble with the bot approach is that it isn't easy to automatically find all the related submissions; especially if you want to filter out the uninteresting threads.
