In the OS (I have no problem with "OS") case, you often see interrupts disabled for a long time because and interrupt would be inconvenient to the OS when it is e.g. pushing onto an artificial XDATA stack.

This is, of course, the great failing of the 8051 unless you are talking about the ones (the few?) with hardware stack space in xdata. I'm not sure how bad the above issue is as I have never used an RTOS on an 8051, but this alone makes me avoid an RTOS for the 8051 because my requirements could not tolerate it. In a sense I'm avoiding them out of ignorance, but I prefer to call it gut instinct. Other applications might have more relaxed timing requirements than mine.

The few apps I do with such relaxed timing requirements hardly merit the time to put the RTOS in, these are 1-day type projects. If I was using an 8051 with a hardware stack in xdata AND there was a mature RTOS that supported it, then I might go that route. I certainly would for most Freescale/Motorola products (save the itty bitty micros).

So perhaps to make Erik's point, no matter how roundabout and verbose, the 8051 RTOS's have such poor context switch time (due to simulated stack in xdata space that any substantial 8051 application would probably require) that they fail to meet most serious user's requirements. This does not apply to other architectures more generous with stack space.

Are we getting closer to a resolution?

GB