I believe that correct terminology for cooperative threads is ?fibers?. However, fibering is essentially a variety of threading, so I don?t think it is totally wrong to use that word. Fibering is like a thread, but with the deteminism of cooperative context switching.

In the single processor world, there are preemptive and cooperative threading models, just as there is preemptive and cooperaitive multitasking. The main difference between multithreading and multitasking is that threads share memory and processes don?t. The context switching mechanism does not have to be preemptive. Cooperative multitasking was used for Windows 3.x, and consequently many see this context switching technique as inferior to preemptive, however I believe this is a bit naive. Cooperative context switching fails to solve certain problems, but as with most things in computer science, a technique may fail in one situation and be perfect in another. In particular cooperative context switching fails to force programs to relinquish control in environments where one program may take control (get stuck) causing the whole computer to be stuck. It creates a computational injustice, all processes suffer, if one doesn?t play nice. This is very annoying a single user computer, but completely unacceptable in a multi-user environment. To solve this problem required preemptive context switching, and there was no sense in OS makers only applying this to process switching and not thread switching.

However, JavaScript runs in a single user AND single process environment. If multiple JavaScript threads can be triggered by developer, a developer can be fully confident that he is responsible for them playing fairly. If one fails he is responsible for fixing it. Therefore, the programmatic advantages of threading can be acheived with the determinism of cooperation.

With the advent of multicore technology more must be considered, plain fibering does not facilitate taking advantage of the different cores. However, I think SMT could be a real bonus here, because it allows one to effectively simulate the determinism of fibering with fibers/threads running on multiple cores.