A minimalist telnet BBS
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
m4
.gitignore
LICENSE
Makefile.am
README.md
app.cpp
configure.ac
db.cpp
db.hpp
editor.cpp
editor.hpp
flymake.mk
form.cpp
form.hpp
login.cpp
login.hpp
mainmenu.cpp
mainmenu.hpp
message.cpp
message.hpp
msglist.cpp
msglist.hpp
pfile.cpp
pfile.hpp
prop.cpp
prop.hpp
pupper.sin
pupper.xin
puppermsg.txt
quotes.txt
screens.txt
sql_password.txt
topic.cpp
topic.hpp
topicmnu.cpp
topicmnu.hpp
upform.cpp
upform.hpp
viewer.cpp
viewer.hpp

README.md

pupperbbs

This is a minimalist telnet BBS.

It is pre-alpha, does not function.

I am old enough to remember BBS via dialup modem. That world still exists out there, but, I don't have a modem any more. But, there is such a thing as a BBS over Telnet. I thought I would write one.

When complete, this should allow one to post to a message board, and to download files via ZMODEM.

License

Libzmodem is GPL2+.

Ncurses is BSD.

The libmysqld client library is LGPL.

So the resultant program is GPL2+.

THE FINISHING THINGS MANIFESTO

  • I will write as little code as possible.
  • I will sensibly and liberally use common libraries and components.
  • I will read and completely understand common libraries I plan to use before I start coding.
  • I will adapt my design to the strengths and constraints created by the libraries and components I use, instead of wrangling them into matching my vision.
  • I will not concern myself with portability until after the program is complete.
  • I will not concern myself about my program being hard to build or stage until after it is complete.
  • I will just accept that GNU Autotools for GNU/Linux and MSBuild on Windows are the best available build tools.

Pupper's internal databases

When thinking about how to set up the databases, I am using as inspiration a much older BBS called Puppy BBS.

The global state

Most of the global state lives for Puppy BBS lived in struct _pup. This is initialized by an INI file, and the INI file is overwritten when values like top or tries change. An INI file that gets modified by the program itself is unfortunate.

Variable Format Init Description
callers int32_t persistent running total of number of callers to the system
quote_pos int32_t persistent current position in quotes file
id _node unused a Fidonet node id
nlimit int16_t INI set max connection length in minutes
klimit int16_t unused max kB download per connection
top int16_t persistent ID of top of circular message buffer
msgnbr int16_t ? current highest message number
messages int16_t INI set total number of messages allowed
msgsize int16_t INI set max chars per message
topic _topic[16] INI set names a descriptions of topics
maxbaud int16_t INI set max baud rate of a connection
mdmstr char[80] INI set initialization string for the modem
cd_bit uint16_t INI set modem Carrier Detect mask
iodev int16_t INI set the COM port number
tries int16_t unused FidoNet dial attempts w/o connects
connects int16_t unused FidoNet dial attempts w/ connects
sched _sched[35] unused Event table for periodic events
filepref char[80] INI set path of upload/download files

This is a mix of constant and persistent variable data. For, Pupper, this might be modified as follows.

The INI constants

Variable Format Init Description
nlimit int16_t INI set max connection length in minutes
klimit int16_t INI set max kB download per connection
messages int16_t INI set total number of messages allowed
msgsize int16_t INI set max chars per message
topic _topic[16] INI set names and descriptions of topics
filepref char[80] INI set path of upload/download files
sqlname char[] INI set username for the MySQL database
sqlpass char[] INI set password for the MySQL database

The persistent data

Variable Format Init Description
callers INT persistent running total of number of callers to the system
quote_pos INT persistent current position in quotes file

Quotes

The quotes database for Puppy BBS was the QUOTES.PUP file. The quotes are stored as variable length plain text, and are separated from one another by a blank line.

For Pupper, I could just store them in the database as a simple map.

Variable Format Init Description
id INT
quote TEXT

Topics

The 16 topics are stored as

Variable Format Init Description
name char[8] INI set short topic name
desc char[24] INI set long topic description

The caller database

In the Puppy BBS the caller was stored as

Variable Format Init Description
name char[36] user name
date uint16_t[16] MS-DOS dates of most recent message read in each topic
topic uint16_t number 1 to 16 of last topic selected
lines uint8_t preferred number of screen lines
cols uint8_t preferred number of screen cols
calls uint16_t number of calls by this user
extra uint16_t unused

The number of callers is limited to 25 in PUP.SET or 100 in defaults

In Pupper, we'll query rows/cols from the Terminfo database.

In Pupper's SQL, this could be

Variable Format Init Description
CallerID INT AUTO_INCREMENT Primary key
name CHAR[36] user name
date DATETIME[16] per-topic dates of most recent message read
topic TINYINT 1 to 16 last topic selected
calls INT number of calls

Where CallerID is the primary key.

The messages database

The in Puppy BBS, the message header was

Variable Format Init Description
from char[36] Name of author
to char[36] Name of recipient
subj char[36] Topic of message
date uint16_t MS-DOS 16-bit date
time uint16_t MS-DOS 16-bit time
extra uint16_t unused
attr uint16_t unused bitmask
topic uint16_t topics
topic_map uint16_t unused ?
message char[2048] The text of the message

The message size of 2048 was probably chosen as around 80 * 25.

The maximum number of messages is limited to 10 by 2560 in PUP.SET, or 50 by 2048 in defaults, with old ones expiring, so the message database is limited to 25k in PUP.SET or 100kB max in default. The 25k was probably for MS-DOS memory handling or floppy disk storage.

For Pupper, in SQL this could be rendered as

Variable Format Init Description
msgid INT UNSIGNED AUTO_INCREMENT the primary key
from CHAR[36] Name of author
to CHAR[36] Name of recipient
subj CHAR[36] Topic of message
timestamp DATETIME MySQL Date/Time
topic TINYINT topics
message TEXT The text of the message

The files database

In Puppy BBS, all downloadable files were stored on the local file system in a single folder. There was a file "FILES.PUP" that held a database of files that could be downloaded.

The "FILES.PUP" file format was

  • filename: char[13]
  • description: char[40]

If FILENAME began with a space or hyphen then description is just a comment.

If FILENAME began with ^Z or @, that was the end of the database.

Other part of the files 'database' comes from the filesystem itself, including creation time and file size.

On Making a Telnet BBS

So original PuppyBBS was probably a MS-DOS program. From the original code, it looks like it started up as a foreground application where the sysop could interact with it, and it also listened to a single modem port.

There are a couple of routes to making this into something I could use.

  • original MS-DOS application with one serial-port-attached modem
  • inetd-style application on my Fedora GNU/Linux server
  • multi-threaded server on my Fedora GNU/Linux server
  • service on my Microsoft Windows 10 box

original MS-DOS application with one serial-port-attached modem

The problem here is testing out a modem, and also that no one will ever call me. Kind of pointless, but, awesome.

The original Puppy BBS ha the sysop-and-user style, with sysop at the console and a user on the modem.

inetd-style application on my Fedora Server GNU/Linux box

This is obviously the simplest. With each new client connection, a new instance of the application is spawned. The TCP I/O is converted to stdin/stdout I/O by the inetd (or systemd) server.

The challenge here is that there needs to be file locking of the database files, since multiple instantiations of the PupperBBS executable may be accessing the files at the same time. Basically, I should use a proper database, instead of ad hoc files, unless I want to write proper file locking, which is a pain on GNU systems.

Info on making systemd to act like an inetd spawner. http://0pointer.de/blog/projects/inetd.html

multi-threaded server on my Fedora Server GNU/Linux box

Now we get a bit more complicated. Multiple threads and multiple connections in a single application.

For info on making a daemon http://0pointer.de/public/systemd-man/daemon.html

For info on making a server application, I can look back at the lessons learned from GNU serveez.

service on Windows 10

This is new to me. There is info about writing services in C++ at MSDN Windows 10 services

It seems a bit like coding up an old school DLL back in the day.

This is also pointless, since my Windows 10 box is not exposed to the world, but it was fun to read about.

Required algorithms

Terminfo and Terminals and Capabilities

According to this pretty bad Wikipedia page, PuTTY and Konsole can do TERM=xterm, and gnome-terminal can do TERM=xterm-256colors.

The TERM=xterm terminfo lists 8 colors and 64 color pairs, with no color modification. Lame. Whatever. In the spirit of the FINISHING THINGS MANIFESTO, I will not try to do more than that.

Hmm. Function keys seem pretty dodgy. The old xterm-r6 used a different and incompatible set of function keys.

If I ever decide to get rid of function keys, consider using TERM=xterm-basic which is basically just xterm without function keys

Function keys

If I use TERM=xterm I get all the function keys.

What have people used function key for, in the past?

F1 help

F2 rename file, cut/copy/paste column, rename symbol, split

F3 search, repeat last command, modify capitalization, end

F4 close program, move cursor to address bar, repeat last command, return

F5 refresh, start slide show, open find/replace dialog, start, stop, find

F6 jump from desktop to taskbar, focus on address bar, jump between menu items and workspace, change, next subwindow

F7 turn on caret browsing, delete current row, up

F8 enter start menu, edit cell, goto next warning, down

F9 compile and run, edit attributes, toggle breakpoint, swap

F10 jump to menu, update data, step over, left, first menu

F11 full screen mode, add new sheet or macro, fill column, right

F12 save as, go to definition, retreive

What are my default function keys, then?

F1 always unused. It is help on Gnome Terminal. useless

F3 search

F4 back or quit action

F8 jump to main menu

F9 submit

F11 unused. It is fullscreen on Gnome Terminal.

The user theory of fixed function keys: fixed function keys are for functions that are continuously available. Fixed function keys always require only one keypress to accomplish a function. Asignments are displayed at all times. Fixed function keys need to indicate an operation immediately after a keypress.

Getting user's terminal size

In original Puppy, the user is queried for the number of rows and columns his terminal supports. In Pupper we're going to ask for the terminal instead. The default terminal size is in the terminfo database.

Note that for ncurses, we need use_env(FALSE) and use_tioctl(FALSE) to be called before any ncurses stuff, so that ncurses uses the number of rows and columns in the terminfo database, instead of trying to query it from environment variables or ioctl calls. Those aren't relevant here because we're running on a client's terminal over telnet.

Replaceable Code

  • ascii.h has #ifdefs like 8 = BS, 9 = TAB
  • edit.c is an interactive console editor. Ncurses Forms also has an editor.
  • quote.c has a date/time parser
  • there is initialization data, which could become an ini file, registry, or database. Boost::property_tree?
  • there is a message file-based database, which could become a standard database. MariaDB
  • the non-portable serial port code could become ASIO to make it both portable serial port and TCP
  • there is a lot of parsing and string handling. Boost::string_algo
  • there is a lot of filename handling.
  • xmodem I/O. https://github.com/caseykelso/xmodem
  • there is a task scheduler. This could also be Boost::ASIO
  • what is the best c++ method to parse user input? getline and regex? Is there something easier than regex?
  • download() searches files by regex.

Libraries