The KeyKOS Nanokernel Architecture (1992)

cis.upenn.edu

77 points·the_why_of_y··10 comments

Predecessor of CoyoteOS and its famous BitC language https://en.wikipedia.org/wiki/Coyotos

Famous? I was on all the mailing lists... there weren't many participants.

The way ahead, at least in the short term, seems to be containerized apps on top of Linux, as with Sandstorm:

I don't know if we'll ever get back to the original vision of very fine-grained apps, with object-cap security baked into the language itself. It would be nice... that would provide another level of security beyond what any OS can provide.

Ha, well I wan't on the ML, and ok not mainstream famous but https://www.google.com/search?q=ltu+bitc yields a lot of discussion about it.

Here are some parts of the KeyKOS source code: http://css.csail.mit.edu/6.858/2011/readings/keykos/

I wrote many of the Keykos papers and can answer most questions you might have. For instance, can sandstorm do confinement as described at http://cap-lore.com/CapTheory/Confine/ ?

This is an important paper and should be read by anyone contemplating the design of secure operating systems.

I agree. It's still more secure by design than any server OS in production today.

Excellent system that exemplified effective use of the capability model. Anyone aiming to learn or develop secure OS's should read on this plus look up the successor for x86, EROS by Shapiro. EROS had a secure kernel, persistance, robust network stack, and more secure GUI. Worth building on.

Note: Look up Coyotos kernel as they mention a few EROS attributes you might improve or change.

Cached version if you're having trouble: https://webcache.googleusercontent.com/search?q=cache:MXidqS...

Proper archive version: https://archive.is/nlKS

Reply on news.ycombinator.com