What ever happened to Tandem and NonStop OS 2094

Of course. I think one become a bit god when one can figure out when to bother with the finer details and when to ignore them. I never could and this is one reason I never agree that I'm a bit god.

Sure. When we talked about black boxes in designs or common converation, it was always implied that complicated things were going on inside it but should not affect the subject of conversation. We could usually call it a bug or serious design flaw if half-way through a project, the innards of a black box started biting the project on the butt.

OTOH, I've seen people get so bogged down in the detail, that no code was written, let alone tested and shipped. We saw happen often during design specs. DECnet design specs suffered from too many details looked at by too many people. Why do you think we have that joke about the committee desiging a horse.

Yes.

In my day, an OS programmer needed to know all of that. That is why one person would be the CPU guy and another person would be the driver guy and another person would be the comm guy.

There is no such thing as perfect. There are always tradeoffs for every choice. For each gain you have to give up something.

That sounds cool.

Optimum is a variable. Access to all paths does not imply fast. It does imply constant access. It's not clear to me that one can have both fast and continuity at the same time. These seem to be what is considered when one has to make a tradeoff decision. I don't know if there are others; I haven't given this one much thought.

You should really watch what is going on with the gamers. They are rediscovering timesharing main frame computing services. Sigh! Caveat: DEC bias alert.

Nope. It would be hardware design that bandaided SMP as an afterthought.

To fix or test anything, requires reproducing the stuff that doesn't seem to work. If two CPUs' instruction end processes require both to depend on each other timings before continuing the next clock cycle of the current instruction, you have serious problems with bug anaylsis. I would baldly state that you don't have an SMP system because the processing is not symmetric and independent of each other.

If each CPU depends on another CPU's interend of an instruction, the system can't be SMP.

Look at gaming. They seem to be encountering the same problems JMF saw before he proposed TOPS-10's SMP design.

BAH