lauren@vortex.UUCP (Lauren Weinstein) (05/10/86)
Just as an aside on the general topic of tandem mode and Unix serial input in general... In my own experience, I've found tandem mode to be largely unusable for most significant applications, especially on smaller machines under any conditions or on large machines under even moderate load. The larger machines often also have problems under light load. Larger machines tend to still lose input data. Smaller machines either lose input data or (even worse, obviously) reliably crash under streams of data at 9600 or sometimes even at 4800. More on that later. Regarding the Telebit modem, it's important to realize that they really did assume that most people would be using PC's running a single program with tightly written interrupt driven code. Their assumption was that for most users, nobody would ever need flow control, since the simple machines would be able to keep up. The model really was a pair of simple PC's talking to each other. The inband (^S/^Q XON/XOFF) handling of the Telebit has problems, possibly related to the buffer management problems that affect other aspects of its performance. Even if Unix didn't have its own serious input problems, the Telebit problems alone would be sufficient to cause concern. The hardware flow control features of the Telebit are better than its inband flow control, but still has problems relating to other firmware issues. Since the turnaround time of the modem has been shortened considerably over its original values, I still strongly suspect that using standard per-packet checking, but with considerably larger packets, will be an important aspect of the best course for using communications protocols with these modems. In any case, I want to emphasize again that the firmware is undergoing rapid change in the Telebit, and that now is not the right time to make decisions or assumptions about the best way to handle the modem. Relatively "minor" firmware changes (from their standpoint) can make major differences in the modem's behavior in these areas. The implementation of a true reverse channel for the modem is a definite possibility, and a variety of other changes may also make a big difference. Since I'll be getting the new firmware releases as they appear at the alpha level, I'll be continuing to track this issue. In the meantime, I strongly recommend sitting back and waiting a while to see what transpires before putting too much effort into dealing with this particular type of modem technology at this time. --Lauren--