On 8/2/23 03:31, D. Hugh Redelmeier via talk wrote:
| From: Anthony de Boer via talk <[email protected]>
| The FTDI USB-Serial adapters were very good, but then the market got flooded
| with cheap PL2303-based ones that are horrible, losing whole swathes of text
| during fast screen updates. They're generally only okay for slow traffic.
What does that mean exactly? The serial-to-USB dongles are not
connected to the screen so "fast screen updates" isn't exactly a
dongle function.
Are you saying that they get over-run during fast or sustained
transmission? That sounds like a driver problem.
The usual situation was using the dongle in the serial port of a device
that was about to start up and output several kilobytes of boot messages
to its serial console at 57600 baud, using minicom in an xterm on my
Linux workstation and the USB-Serial dongle to watch what was happening,
and realizing chunks were missing from the story the device was trying
to tell me.
I think that if it was a driver problem our Linux community would have
jumped on it, especially for something as ubiquitous as the PL2303, and
when I looked into it the news was that it was a chip problem.
Even in the old days, overrun was too easy with some serial input ports.
Original UARTs were double buffered. 8250's like in the original PC had
deeper buffers. If you don't have deeper buffers, you need the OS to have
very quick real-time response to serial port interrupts.
Old-timer stories -- feel free to ignore:
My Kaypro II could handle 9600 terminal emulation only because I rewrote
parts of the firmware (CP/M BIOS in EPROM) for faster screen output and
wrote my own terminal emulator to use serial port interrupts.
The NABU 1600 could handle 9600 serial input only because Sheila Crossey
of HCR wrote some code in the UNIX kernel to handle serial input in a
pseudo-DMA way.
I think I've still got a Kaypro II somewhere because I just can't stand
to e-waste such great engineering of yesteryear, especially after
spending quality time with it back then.
The closest I got to the serial ports was writing an xmodem for CP/M to
try and juice a bit more transfer speed out of our hardware, and later
porting that to early MSDOS. That would have been nearly the last
assembler I wrote before gaining access to the new C compilers that were
starting to appear.
A bit later we had Unix servers with dozens of users on serial ports all
talking to a single-core 386 or 68k and we thought these was a good
systems. A big part of it was smart multi-port serial cards that handled
"cooked-mode" line buffering for the users and only interrupted the OS
when there was a whole line of input to process. That would have been
mid-80s into the early 90s.
Anthony
---
Post to this mailing list [email protected]
Unsubscribe from this mailing list https://gtalug.org/mailman/listinfo/talk
