That largely misses (Dijkstra's, originally) point. How control flow is implemented at a low level (e.g. jmp/lngjmp) is a completely separable from how it should be exposed in a language.
Wearing a c programmers' hat you may say "absolutely", a scheme programmers' hat, perhaps "no way". Horses for courses after all.
>>ska+(OP)
That's a more nuanced position than the one that is commonly understood, however. Though, you are always bound to your architectural model, so exploiting it more directly is not innately bad. Layers tend to be leaky.
Wearing a Scheme programmers' hat, call/cc isn't any less of a landmine.