From pfarrell@cs.gmu.edu Tue Jan 26 12:28:05 1993
While serial ports do lose characters, especially if you don't have a 16550afn serial chip, I don't see this as a major hassle. In a pure DOS space, you really arn't likely to lose the characters, and this is the initial space of NUpop. With Windows, you have to learn to play with the priorities to make it work well, or get one of the intellegent serial driver DLLs that make it transparent.
Let me try to put it another way. The higher you drive the DTE rate, the more likely you are to loose characters. At the same time, you start to care less about the (small) protocol overheads involved.
It is possible that the authors of the NUpop document don't worry too much about single character dropouts. There is plenty of redundancy in english. PGP will complain, but I can see retransmitting a message half a dozen times to get it thru cleanly will lose.
But if characters change during transmit, how can you tell that the message wasn't altered by some agent other than the device/driver? Further, if it happens only occasionally, won't you react with mistrust of the original message? "Hey, this message doesn't check with the author's key!" If it happens a lot, aren't you more likely to say, "Well, he must have meant $10,000, not $1000, the serial port must be loosing again." rather than resending some number of times?
I never allow my private key anywhere near a serial port. The public keys are checksummed, so it is easy to see that a character is wrong.
I'd love to be able to use CSLIP. We (a bunch of folks on this campus) have just convinced the admin to allow POP services. It will take a while before we can convince them to allow SLIP, CSLIP, and PPP.
The older I get, the more I understand, "Power to the people." Cheers, Jim