porttiki@cs.Helsinki.FI (Anssi Porttikivi) (11/18/90)
Hello, all you UNIX(tm), AIX(tm) and data communications experts! We have BIG problems runnning AIX PS/2 X.25. Our hardware: IBM PS/2 80-111, 386 20 Mhz, 4 Mb RAM, 110 Mb disk IBM X.25 card Nokia synchronous modem HPY (the local phone company) X.25 socket in the wall Our (buggy) software: IBM AIX PS/2 1.2 AIX PS/2 X.25 Program IBM maintenance update 01.02.0300 Our mission: To set up a public access multi-user USENET site. The telephone company provides a service by which people can call our public number with ordinary modems and the calls are routed and protocol converted to our X.25 address. This is supposed to be a cheap multi-user alternative to multiple telephone lines. Our problems: We have serious problems that prevent any intended use of the software and the IBM staff here are not too helpful, or skilled enough. Problem 1. The X.25 first refused to do anything. It began to initiate sessions after applying update 01.02.0300. However, sessions through X.25 do not work properly: - when editing the command line the "erase" character messes up the display. The edited command line is given to the system OK, but the screen looks like backspace is echoed as <BS><BS><space><BS>. - when running microEmacs (we compiled the sources), characters are not echoed immediately, but seem to be buffered. I can type about 20 characters before they appear on the screen all at once. If I type one character and wait a few seconds that one character eventually appears. The problem doesn't appear to be a slow line or system, since return and microEmacs commands work immediately. If I type return or a few characters and return they appear on the screen immediately. Using some command, Control-A for example, instead of return also gets immediate echo. Our theory is: The X.25 processor tries to off-load processing from the CPU, but instead of being helpful gets in the way. Looks like writing on the command line makes X.25 echo everything typed, but backspace is echoed also by the tty driver as <BS><space><BS>, which in addition to the <BS> echoed by X.25 messes up the screen. When in microEmacs (version 3.10), X.25 handles letters and digits itself, buffering them. When X.25 gets something it does not know what to do with, it gives the buffered input to the CPU. Why does it work differently in microEmacs? As far as we can understand, the X.25 subsystem should immediately pass everything typed to the tty device driver and not buffer anything. Also it should not echo anything. We tried "stty" command, but did not understand what to change. We seem to get error messages trying the "stty" on the X.25 system pseudo tty port device files. Just writing this I received a message from a friend. He says, that the X.25 standard defines a parameter FORWARD which controls the echoing of different cahracter classes. As far as I know this not mentioned in the AIX X.25 documentation? Problem 2. When we initiate a session through X.25, we get modem commands thrown at us by don't know which software, maybe "getty" ? This is very harmful in our case and must be corrected. The commands are those Hayes commands used to initialize a modem for auto-answer and some others. How should I configure the line so that all programs involved believe it is not a modem line? Setting clocal + or - doesn't seem to affect anything and the IBM X.25 software doesn't always allow using the "stty" to set CLOCAL. When a X.25 session ends an error message "Could not set clocal, getty sleeping." is always displayed. Is anybody in the whole world running this kind of system? Can it work? Are we stupid? Help us! -- Anssi Porttikivi Anssi_Porttikivi@cs.helsinki.fi "I am Gaudeamus Data Oy (Inc.) Vuorikatu 5 Finnished telefax (+358-0-) 13114345 00250 HELSINKI, FINLAND now"