News for hackers. Stuff that matters.

A comment on The Taos Operating System (1991). Parent comment

Re:The Taos Operating System (1991) by JoeAltmaier on Tuesday June 30, 2015 @05:26PM (#9808538)
You could imagine an OS with no threads. E.g. dataflow, or message-processors working queues, or something else. Perhaps instead of processes we'd use just virtual machines. Maybe instead of managing memory, we manage only data abstractions e.g. capabilities. There are strong reasons to want safety and security guarantees around the lifetimes of those things.

All of those should properly be managed by a kernel, yes?

[ Parent | on HN ]
Re:The Taos Operating System (1991) by haberman on Tuesday June 30, 2015 @05:47PM (#9808631)
> You could imagine an OS with no threads. E.g. dataflow, or message-processors working queues, or something else.

Sort of like Erlang? Which can run on POSIX.

> Perhaps instead of processes we'd use just virtual machines.

Sort of like the JVM, or Docker, or VirtualBox, or any number of VM technologies that run on POSIX?

> There are strong reasons to want safety and security guarantees around the lifetimes of those things.

Why do any of these things need to live in the kernel to give safety/security guarantees? If anything, code that runs in the kernel is more vulnerable, not less.

btw. I don't think POSIX is the be-all end-all or anything. But if I imagine something better than POSIX, its abstractions are even less opinionated than POSIX, not more.

[ Parent | on HN ]
Re:The Taos Operating System (1991) by pjmlp on Wednesday July 01, 2015 @03:10AM (#9810389)
Getting rid of POSIX is the only way of achieving safer OSes. POSIX is nothing more than the C runtime vs what other languages with richer runtimes offer. They standards body just decided to split the runtime between ANSI and Open Group.

I only care about OS POSIX support every time I have to go down to C or C++. Otherwise the OS can have whatever abstraction model it feels like.

[ Parent | on HN ]
Re:The Taos Operating System (1991) by JoeAltmaier on Tuesday June 30, 2015 @05:54PM (#9808668)
Putting things in a kernel control failure modes, security, performance at the cost of rigid APIs. So I understand wanting to avoid committing to a kernel. But its harder to make strong guarantees without.
[ Parent | on HN ]