Data communications over telegraph circuits 1909

And if many lives depend upon it, or if the consequences are otherwise that serious, then it *is* necessary. Nuclear power plants and 747's are obvious examples. Fault tolerant systems for banking and telecommunications are not quite as obvious, and have some other interesting characteristics that few people are aware of.

But look what happened when the SS7 code had a single malformed for loop, and shutdown half the toll network within minutes...

...

Hence you are pointing out that yes it can tolerate the small nuisance failures, but *not* mbuttive outages.

...

Interesting. But not significant to the point being discussed.

Also interesting is that fault tolerance for banking is defined as preserving the *data* at all costs. The transaction process might fail, but "who owns what" cannot be lost.

The fault tolerance built into telephone switching equipment is at least as significant as that in bank computing, but has a very different objective: preserve the process at all costs, but allow the data to be lost!

Which means that for failures in the telco's computer, your call is not supposed to away. You may not get billed for it though, because if a glitch does occur it might just dissociate the data on when the call started from the data on when the call ended. (Who owns the money counts for banking, but not for telephone calls... and a typical reboot of a telco switch doesn't dump calls but does lose billing data.)

That's too late when the plane is 200 feet off the ground and the autopilot failure mode causes *any* control surface to malfunction by moving in any direction. In essence, during take off and landing a mbuttive computer failure is bane.

Exactly. If it fails to protect, people die. Mbuttive failure is not tolerable.

--