[jvanenk]

I'm reminded of protothreads (http://dunkels.com/adam/pt/). I used protothreads some time ago on an embedded project to implement a (very small) HTTP server in only a few KB or RAM.

[_yosefk]

The trouble with them is they're stackless, which means you can't have arbitrary functions thoughtlessly coded by someone unaware of threading called inside many different protothreads. Coroutines are easier to combine with code not written to account for their existence (of course stack overflow is a big concern though, which can be somewhat mitigated through memory protection depending on the OS and the hardware).

[enqk]

Care to explain further?

In my experience of using protothreads, only the functions using the protothread macros need to be written in a special way:

   - passing a context object
   - being careful of variable initialization/blocks 
     in order not to work on unitialized memory.

while all the other ones can stay plain C functions.

[_yosefk]

Sure, if none of these C functions tries to yield, or calls a library function trying to yield. In fact, the problem isn't that they can't be plain C functions, but that they must be plain C functions :)

Why would functions called by "lightweight threads"/"coroutines"/whatever want to yield? One example is, they want to wait for an event - an I/O request completion or a computation offloaded to an accelerator or whatever - without having to return all the way back to the event loop.

[enqk]

I must admit that this property attracted me to protothreads, since there is a clear demarcation between coroutines (through the presence of a protothread argument) and plain C functions.