Last update 01-Mar-2007
Running a BBS under Unix.
Everyone who has been running a (single line) BBS under DOS until now will need to understand that running a BBS under Unix (or any other multitasking os) is completly different of what you are used to. Under DOS things were quite simple, from AUTOEXEC.BAT you started a new .BAT file that would run forever and started all needed programs after each other. The programs that where started depended on the errorlevel of the previous program. Only one program could run at the same time.
People who had previous run a BBS on another multitasking os, or were running a BBS on a small lan with a fileserver and workstations for each line, are already more used to the idea of running more programs at the same time, and to "signal" what to do next with semafore files.
The Unix aproach is more or less the same, but there are more differences. The main difference is that there is no mailer connected with the modem waiting for a call, instead there is a getty process watching your modem(s). And for internet calls, these are started via (x)inetd using standard *nix methods. Another big difference is that you don't see what's happening, there is no screen with the mailer or bbs picture on it. All programs run in the background. If you don't like that, stop now and go back to your old DOS bbs. It's just the way everything is done.
Programs that must start at specific times (events in DOS), are started from cron, this is the event scheduler for Unix. With this program maintenance can be started, polls created etc. For starting programs when they are needed there is a taskmanager loaded at system bootup. This taskmanager "watches" the semafore directory of the bbs and will start what is needed. It will also watch the mailer outbound to see if there are nodes to call.
Waiting for a call .....
Under Unix this is done with the mgetty program, this is the process that is connected with each modem (or ISDN adapter) and waits for a call. The mgetty program (written by Gert Doering, firstname.lastname@example.org) will detect the call, and find out what or who did make the call. It can detect incoming humans who want a login prompt, PPP calls from users who want to make a PPP connection (browsing your BBS whith netscape for example), A fax machine trying to deliver a fax and finally a mailer trying to establish an EMSI, FSC-0006 or FSC-0001 session. The mgetty program is responsible for starting the right client programs. How to do this is explained in the installation manuals, but be sure to compile it with Fido and PPP support.
For internet calls, via (x)inetd mbsebbs can accept calls from connections on ports. For example if a connection comes in on port 24554 the mailer is started in binkp mode.
A Human is calling.
This could be a bbs user. For each user to login to your bbs there is a unix account. They automatic create such an account the first time they login with the bbs account. During the creation of their account the shell that is installed for there account is the mbsebbs binary, so that's the only thing that they get if they call in. When they logout the bbs, or drop carrier etc, the session is ended and mgetty takes over the line again. Note that they never can get a Unix shell unless you install a door in the bbs that calls a shell for them.
Because for dialin with mgetty the mblogin program is used and not the normal login program, only bbs users can login. All other users including root are denied access. You might consider installing SSH on your system for remote maintenance.
It is also possible to install the mblogin program as the login program for telnet users if you are directly connected to the internet.
A PPP call is detected.
Installing a PPP server on your system is beyound the scope of this project. However if you did install it, users can login your bbs with their favourite browser and use your bbs. Note that the necessary tools to automatic create newsgroups don't exist at this time. With the proper setup you can automatic create and maintain html pages for the file areas.
A mailer call is detected.
If a mailer is detected by mgetty, the mbcico program is started and will take over from mgetty. It will establish a mail session with the caller and the mail and or files will be exchanged just like any DOS mailer would do. After the call, mbcico will hangup and mgetty will take control of your modem again. If there is any mail received, mbcico will place the semafore mailin so that another process can take care of the received mail.
There is mail in the inbound
As I said before, if the mailin semafore is present, the task manager will then start the mbfido program that will toss the mail, process any files received and if necessary it will create other semafore's for example to link the message bases, start the nodelist compiler etc. Note that this can be done while there may be a new mailsession going on, a bbs user is online, it doesn't matter. Processing mail and files can be done real multitasking without any damage to other processes.
It's time to poll a node
At the time that you whish to poll a node, let cron create "poll" requests. When a poll is created, the semafore scanout is also created. The taskmanager will then start mbcico at regular intervals so that mail will get out. If there is no more mail to send, the scanout semafore is removed. If a timeslot ends, you can just remove the "poll" requests that didn't succeed.
It's Zone Mail Hour, so now what
Relax, if you have netmail ready for nodes the mailer script will try to send these mails to those nodes. If it was crash mail, and the destination was a non CM node, the mailer will try to send those mails too. Note that other crashmails are send anytime. Also note that packed mail and files are not send during ZMH. If a node calls you during ZMH he will get everything that's waiting, including packed mail and files. The task manager (more on that later) calculates the Zone Mail Hour from UTC time, you don't have to change anything for summer- and wintertime. The taskmanager knows about all zone mail hours of all Fidonet zones (1..6). For othernets, zone mail hour is taken from the setup as default value. The taskmanager will also handle the mail for nodes with the Txx flag (FSC-0062) on the right times.
This is started by cron jobs. There is no need to take your bbs lines down during maintenance, you can do it any time of the day. I have made several scripts for this, daily, weekly and monthly.
How about system load
Because GNU/Linux is a 32 bit os, not bothered with a graphical user interface (unless you install it), it has all the time in the world to serve your bbs programs. Background programs are build to release time to the Unix os, they don't need to run fast because it's background processing. The bbs and the mailer, have a low server load although there is no timerelease build in. Only the bbs has some short moments when it needs a lot of your system, for example when a user logs in and scans for new mail. The bbs I run is a 486-DX4 100 MHz, 20 MB ram, with 2 analogue lines, this seems to work fine. When this system's MOBO died, I used a 386DX33 for several months with 20 MB ram, and the only thing users ever noticed was that scanning for new mail was slower. I think this is the slowest harware that will work. However, you must always use 16550A uarts for the COM ports. For best performance use SCSI disks. I noticed that old 5"FH SCSI disks perform better for bbs usage then modern EIDE disks. This is probably caused by the fact that the kernel needs more time for the cheap IDE bus. If you want to use X11 on your bbs, you need more ram and a faster CPU or a separate machine via a lan and export the display to that machine. Currently the smallest amount of memory you need is 36 MB, and that amount is only needed if you use a virus scanner which needs more memory then mbse.