wht@n4hgf.Mt-Park.GA.US (Warren Tucker) (12/16/90)
In article <1634@boink.UUCP> harald@boink.UUCP (Harald Milne) writes: >Conclusion: SIO is broken, and broken bad enough to crash your system. >Don't use SIO with dumb serial cards. Hmmmm.... I do asynchronous communications like most people breathe... pretty often. I run two high speed modems (a Telebit T2500 and a Microcomm 9624), an X-10 controller, another PC, an AMTOR packet controller, a Wyse-60 terminal, an HF radio controller, a 25MHz-1.2GHz receiver controller, a serial mouse and a Ku-band VSAT, at speeds from 2400 to 38.4 and duty cycles from 0 to 100%, all with dumb cards (Digiboard COM-8 and standard COM1), 16450s and stock SIO. I have a bit of trouble receiving streaming 38.4 with the 450s during disk activity, but since I do async protocol development, I derive a perverse sense of pleasure at being able to generate errors merely by doing a find / :-). I have little or no problem with 9600 baud or below during any system activity (my compressed news feed comes over the T2500 at 19.2 with 1400 ch/sec throughput, insignificant error rate). I plan to add 550s to some of the ports, but so far haven't really felt hurt enough to bump it very high on the priority list. Likewise, I have studied the wonderful FAS driver and keep thinking over putting it in, but I don't need it except for custom EIA signal diddling I'd like to get around to. Of course, this is on a 20 MHz 386, but I had an 8MHz 286 for three years and did abundant amounts of async at 9600 (maybe 3 billion characters/month) with 2.0.6, 2.1.1 and 2.3.1 XENIX 286, with good luck. I have had zero PANICs in six years of 24-hour per day, 50% average utilization operation on a Compaq Deskpro 286/8 and 386/20. The purpose of my verbosity isn't just to sound like Scotty on a tour for the Admiralty, but to avoid you starting another suburban legend. I'm pretty fond of SCO's SIO, much more tolerant and adaptive than any other I have seen. Maybe I just hold my face right, but it -is- in a smile. Smart serial cards are OK, but it's nice to be able to stick with the straight and narrow: most smart cards don't use 8250 lookalikes. I've programmed 8274s, 8530s, Z80 SIO, 6850 ACIAs, 68561 DUSCCs, and many others, even the wretched 68901. The 8250 clan ain't the best, ain't the worst, but there are more than a few in use. (Hint: Some enterprising soul should put a pair of DUSCCs on an AT card. Those devices do everything but laundry!) ---------------------------------------------------------------------------- Warren Tucker emory!n4hgf!wht or wht@n4hgf.Mt-Park.GA.US "I was 35 years old before I knew a pie was meant to be eaten." - Moe Howard
daveh@marob.uucp (Dave Hammond) (12/18/90)
wht@n4hgf.Mt-Park.GA.US (Warren Tucker) writes: harald@boink.UUCP (Harald Milne) writes: >>Conclusion: SIO is broken, and broken bad enough to crash your system. >>Don't use SIO with dumb serial cards. > >Hmmmm.... I do asynchronous communications like most people >breathe... pretty often. [...] Likewise. I have been putting SCO on 286 machines (from ATs to no-names) since 2.1.3 (circa 1985) and have never had a panic caused by SIO. I've used generic single-port "COM2:" cards and dumb 4-port cards from Arnet, DigiBoard and AST. These have been running in production machines without incident for several years. My shop still installs at least one 286 Xenix machine per month (running 2.3.2). We would not be in business if the SIO caused machines to crash and burn. I suggest that harold@boink investigate possible hardware problems. -- Dave Hammond daveh@marob.masa.com