secrist@msdoa1.DEC.COM (Richard C. Secrist, KXO SWS) (01/06/88)
I forget all my sources but when I "switched to MEX" I set out to resolve your same problem myself. Without reviewing all of my tests, my results were as follows. I talked to people at PCPI and Laing Electronics (who sell and repair PCPI stuff) who both said it was tied to the video card in use; checking other sources the consensus was that the Ultraterm card worked best and solved it through like 9600 bps. I only half bought this story, becuase the same serial card and video hardware under "native" 6502 modes like Apple DOS or PASCAL seemed to work pretty well - better than on the PCPI anyway. I wrote a little FORTH dumb terminal program in 1K in native mode and it didn't but a couple here and there. ASCII Express under DOS 3.3 never dropped anything - but it does interrupt I/O to RAM and displays it later. So I decided the key on the Apple was buffering the characters as they come in BEFORE displaying them, since the display is so miserably slow. Then Apple came out with the "enhanced //e" with completely re-written 80-column firmware, and lo and behold my little 1K FORTH glass TTY worked fine. I later got a VT-100 emulator called DCOM v3.3 that works fine doing ANSI escape sequences at 2400 baud using the buffered technique WITHOUT interrupts. So on my "to do" list is a task to write a 6502-resident driver that buffers character I/O that the PCPI side can get characters from over the comparatively slow parallel CPU port that uses Apple's 6502 to do all I/O anyway. By the same token, the Ultraterm card is supposed to fix this too. So here was my solution: since I wanted VT-100 or at least VT-52 emulation and MEX didn't fit that bill I really only needed a usable glass TTY to fire up XMODEM and do directories and so forth. My workaround was to tell the host operating system to add a bunch of fill characters and that works fine for what I wanted out of it. Under VAX/VMS I did a "$ SET TERM/LA36/CRFILL=9/LFFILL=9" ...note that since v4.0 of VMS "=9" is "class nine" and not "nine characters". Previously (v3.n) I think I used (3) CR fills and (7) LF fills or vice-versa. If you use VMS and think you're having flow control problems VAXen don't normally have flow-control on in both directions - do a "$ SET TERM/HOSTSYNC/TTSYNC" to fix that. When I need real terminal emulation, I go back to native Apple modes, and that wins through 2400 baud on an "enhanced //e." Hope that helps... Has anybody 1) seen KERMIT for the PCPI, 2) seen a MEX overlay for a DEC "Scholar" or DF03 modem ? rcs