@JackEric @davidallengreen Yes. We can say the parts which ran did so correctly after error correction. ECC, for example, saves butts constantly. Sure, a redundant system worked. But redundancy was there b/c we knew failure was inherent in the design. Now add in software. There is no test for computers running correctly, but give me one to beat on w/the right equipment & I'll likely find errors. To say they work correctly by default is just crazy talk to crazy talkin' me, lol.
@stevesplace @JackEric @davidallengreen
Thankfully none of the Horizon system ran on Sun SuperSPARC hardware, a platform designed with a fundamental flaw. No ECC on databuses. Sun wrote a patch to determine if an error occurred in kernel code, if that was the case the crashed, if user code it just kept going.
Horizon did migrate to Solaris but on Fujitsu’s own SPARC based design which exploited mainframe principles.
That was just the hardware.
Horizon didn’t involve formal methods, in the early phases (ever?), everyone was learning. Corporate/institutional ‘knowledge’ was migrating from an ICL-VME mainframe mindset to NT as an enterprise platform. NT had many gaps as an enterprise platform in the early development of Horizon.
To make the assumption, a computer should be regarded as operating correctly, is deeply flawed. Attempting to apply caveats would be similarly flawed.
