Rayla (rayla.delete@this.example.com) on June 29, 2022 2:14 pm wrote:
> Kester L (nobody.delete@this.nothing.com) on June 29, 2022 1:49 pm wrote:
> > https://queue.acm.org/detail.cfm?id=3534854
> >
> >

> > The linear address space as a concept is unsafe at any speed, and it badly needs mandatory CHERI
> > seat belts. But even better would be to get rid of linear address spaces entirely and go back to
> > the future, as successfully implemented in the Rational R1000 computer 30-plus years ago.
> >

> >
> > Your thoughts on this article? I was under the impression that a lot of the 80s attempts
> > at capability machines (or really, anything that wasn't trying to be a glorified PDP-11)
> > floundered because of performance and cost issues (i.e. the Intel i432).
> >
> >
>
> 432 was a seriously substandard implementation with good ideas, and has been getting used to proclaim that descriptors/caps
> are fundamentally broken ever seince. Meanwhile, Burroughs Large Systems (still 1000+ sites, almost all very
> large companies or governments) and IBM i (over 100k sites, occasionally even greenfield customers) both quietly
> zoom along running production workloads today, but are ignored because they aren't UNIX.
>
> The Colwell paper on 432 is an interesting read, and heavily implies
> that the 432 could have been greatly improved with only minor tweaks.

So basically the only real reason why PDP-11-esqe systems with flat memory spaces proliferated is basically because of the old worse-is-better effect where an easier-to-initially-implement system dominated because it's easier to get started with it at better performance given the limitations of the technology of the era (i.e. Unix).