README

=====================================================================
  libtelnet - TELNET protocol handling library
=====================================================================

 http://github.com/elanthis/libtelnet

 Sean Middleditch
 sean@sourcemud.org

---------------------------------------------------------------------
The author or authors of this code dedicate any and all copyright
interest in this code to the public domain. We make this dedication
for the benefit of the public at large and to the detriment of our
heirs and successors. We intend this dedication to be an overt act of
relinquishment in perpetuity of all present and future rights to this
code under copyright law. 
---------------------------------------------------------------------

*** TODO ***

 - automatic MCCP2 handling (controllable by host app)
 ? ZMP parsing
 ? MSSP parsing
 ? ENVIRON/NEW-ENVIRON parsing
 ? telnet-status testing tool

I. INTRODUCTION
=====================================================================

libtelnet provides safe and correct handling of the core TELNET
protocol.  In addition to the base TELNET protocol, libtelnet also
implements the Q method of TELNET option negotiation.  libtelnet can
be used for writing servers, clients, or proxies.

For more information on the TELNET protocol, see:

 http://www.faqs.org/rfcs/rfc854.html
 http://www.faqs.org/rfcs/rfc1143.html

II. LIBTELNET API
=====================================================================

The libtelnet API contains several distinct parts.  The first part is
the basic initialization and deinitialization routines.  The second
part is a single function for pushing received data into the telnet
processor.  The third part is the telnet_send_*() functions, which
generate TELNET commands and ensure data is properly formatted before
sending over the wire.  The final part is the event handler
interface.

IIa. Initialization

 struct telnet_t;
   This structure represents the state of the TELNET protocol for a
   single connection.  Each connection utilizing TELNET must have its
   own telnet_t structure, which is passed to all libtelnet API
   calls.

 void telnet_init(telnet_t *telnet, telnet_event_handler_t handler,
     unsigned char flags, void *user_data);
   The telnet_init() function is responsible for initializing the
   data in a telnet_t structure.  It must be called immediately after
   establishing a connection and before any other libtelnet API calls
   are made.

   The handler parameter must be a function matching the
   telnet_event_handler_t definition.  More information about events
   can be found in section IId.

   The user_data parameter is passed to the event handler whenver it
   is invoked.  This will usually be a structure container
   information about the connection, including a socket descriptor
   for implementing TELNET_EV_SEND event handling.

   The flags parameter can be any of the following flag constants
   bit-or'd together, or 0 to leave all options disabled.

    TELNET_FLAG_PROXY
      Operate in proxy mode.  This disables the RFC1143 support and
      enables automatic detection of COMPRESS2 streams.
 
 boid telnet_free(telnet_t *telnet);
   Releases any internal memory allocated by libtelnet.  This must be
   called whenever a connection is closed, or you will incur memory
   leaks.

IIb. Receiving Data

 void telnet_push(telnet_t *telnet,
     const char *buffer, unsigned int size, void *user_data);
   When your application receives data over the socket from the
   remote end, it must pass the received bytes into this function.

   As the TELNET stream is parsed, events will be generated and
   passed to the event handler given to telnet_init().  Of particular
   interest for data receiving is the TELNET_EV_DATA event, which is
   triggered for any regular data such as user input or server
   process output.

IIc. Sending Data

 All of the telnet_send_*() functions will invoke the TELNET_EV_SEND
 event.

 Note: it is very important that ALL data sent to the remote end of
 the connection be passed through libtelnet.  All user input or
 process output that you wish to send over the wire should be given
 to one of the following functions.  Do NOT send or buffer
 unprocessed output data directly!

 void telnet_send_command(telnet_t *telnet, unsigned char cmd);
   Sends a single "simple" TELNET command, such as the GO-AHEAD
   commands (255 249).

 void telnet_send_telopt(telnet_t *telnet, unsigned char cmd,
     unsigned char telopt);
   Sends a TELNET command with an option code following.  This is
   only useful for the WILL, WONT, DO, DONT, and SB commands.

 void telnet_send_negotiate(telnet_t *telnet, unsigned char cmd,
     unsigned char opt);
   Sends a TELNET negotiation command.  The cmd parameter must be one
   of TELNET_WILL, TELNET_DONT, TELNET_DO, or TELNET_DONT.  The opt
   parameter is the option to negotiate.

   Unless in PROXY mode, the RFC1143 support may delay or ellide the
   request entirely, as appropriate.  It will ignore duplicate
   invocations, such as asking for WILL NAWS when NAWS is already on
   or is currently awaiting response from the remote end.

 void telnet_send_data(telnet_t *telnet, const char *buffer,
     unsigned int size);
   Sends raw data, which would be either the process output from a
   server or the user input from a client.

   For sending regular text is may be more convenient to use
   telnet_printf().

 void telnet_send_subnegotiation(telnet_t *telnet,
     unsigned char telopt, const char *buffer, unsigned int size);
   Sends a TELNET sub-negotiation command.  The telopt parameter is
   the sub-negotiation option.

   Note that the above function is just a shorthand for:
    telnet_send_telopt(telnet, TELNET_SB, telopt);
    telnet_send_data(telnet, buffer, size);
    telnet_send_command(telnet, TELNET_SE);

   For some subnegotiations that involve a lot of complex formatted
   data to be sent, it may be easier to manually send the SB telopt
   header and SE footer around mulitple calls to send_data.

   NOTE: telnet_send_subnegotiation() does have special behavior in
   PROXY mode, as in that mode this function will automatically
   detect the COMPRESS2 marker and enable zlib compression.

 int telnet_printf(telnet_t *telnet, const char *fmt, ...);
  This functions very similarly to fprintf, except that output is
  sent through libtelnet for processing.  IAC bytes are properly
  escaped, C newlines (\n) are translated into CR LF, and C carriage
  returns (\r) are translated into CR NUL, all as required by
  RFC854.  The return code is the length of the formatted text.

  NOTE: due to an internal implementation detail, the maximum
  lenth of the formatted text is 4096 characters.
 
 int telnet_printf2(telnet_t *telnet, const char *fmt, ...);
  Identical to telnet_printf() except that \r and \n are not
  translated.  This should be used if you are attempting to send
  raw data inside a subnegotiation or if you have already manually
  escaped newlines.

IId. Event Handling

 libtelnet relies on an event-handling mechanism for processing the
 parsed TELNET protocol stream as well as for buffering and sending
 output data.

 When you initialize a telnet_t structure with telnet_init() you had
 to pass in an event handler function.  This function must meet the
 following prototype:

  void (telnet_t *telnet, telnet_event_t *event, void *user_data);
 
 The event structure is detailed below.  The user_data value is the
 pointer passed to telnet_init().

  struct telnet_event_t {
    const char *buffer;
    unsigned int size;
    telnet_event_type_t type;
    unsigned char command;
    unsigned char telopt;
    unsigned char accept;
  };
 
 The enumeration values of telnet_event_type_t are described in
 detail below.  Whenever the the event handler is invoked, the
 application must look at the event->type value and do any necessary
 processing.

 The only event that MUST be implemented is TELNET_EV_SEND.  Most
 applications will also always want to implement the event
 TELNET_EV_DATA.

 Here is an example event handler implementation which includes
 handlers for several important events.

  void my_event_handler(telnet_t *telnet, telnet_event_t *ev,
      void *user_data) {
    struct user_info *user = (struct user_info *)user_data;

    switch (ev->type) {
    case TELNET_EV_DATA:
      process_user_input(user, event->buffer, event->size);
      break;
    case TELNET_EV_SEND:
      write_to_descriptor(user, event->buffer, event->size);
      break;
    case TELNET_EV_ERROR:
      fatal_error("TELNET error: %s", event->buffer);
      break;
    }
  }

 TELNET_EV_DATA:
   The DATA event is triggered whenever regular data (not part of any
   special TELNET command) is received.  For a client, this will be
   process output from the server.  For a server, this will be input
   typed by the user.

   The event->buffer value will contain the bytes received and the
   event->size value will contain the number of bytes received.  Note
   that event->buffer is not NUL terminated!

   NOTE: there is no guarantee that user input or server output
   will be received in whole lines.  If you wish to process data
   a line at a time, you are responsible for buffering the data and
   checking for line terminators yourself!
 
 TELNET_EV_SEND:
   This event is sent whenever libtelnet has generated data that must
   be sent over the wire to the remove end.  Generally that means
   calling send() or adding the data to your application's output
   buffer.

   The event->buffer value will contain the bytes to send and the
   event->size value will contain the number of bytes to send.  Note
   that event->buffer is not NUL terminated, and may include NUL
   characters in its data, so always use event->size!

   NOTE: Your SEND event handler must send or buffer the data in
   its raw form as provided by libtelnet.  If you wish to perform
   any kind of preprocessing on data you want to send to the other
 
 TELNET_EV_IAC:
   The IAC event is triggered whenever a simple IAC command is
   received, such as the IAC EOR (end of record, also called go ahead
   or GA) command.

   The command received is in the event->command value.

   The necessary processing depends on the specific commands; see
   the TELNET RFC for more information.
 
 TELNET_EV_WILL:
 TELNET_EV_DO:
   The WILL and DO events are sent when a TELNET negotiation command
   of the same name is received.

   WILL events are sent by the remote end when they wish to be
   allowed to turn an option on on their end, or in confirmation
   after you have sent a DO command to them.

   DO events are sent by the remote end when they wish for you to
   turn on an option on your end, or in confirmation after you have
   sent a WILL command to them.

   In either case, the TELNET option under negotiation will be in
   event->telopt field.

   If you support the option and wish for it to be enabled you must
   set the event->accept field to 1, unless this event is a
   confirmation for a previous WILL/DO command you sent to the remote
   end.  If you do not set event->field to 1 then libtelnet will send
   a rejection command back to the other end.

   libtelnet manages some of the pecularities of negotiation for you.
   For information on libtelnet's negotiation method, see:

    http://www.faqs.org/rfcs/rfc1143.html

   Examples:

    You want remote end to use TTYPE, so you send DO TTYPE.
    Remote accepts and sends WILL TTYPE.

    Remote end wants you to use SGA, so they send DO_SGA.
    You do not support SGA and set event->accept = 0.

    Remote end wants to use ZMP, so they send WILL ZMP.
    You support ZMP, so you set event->accept = 1 and enable
    local ZMP support.

    You want to use MCCP2, so you send WILL COMPRESS2.
    Remote end accepts and sends DO COMPRESS2.

   Note that in PROXY mode libtelnet will do no processing of its
   own for you.

 TELNET_EV_WONT:
 TELNET_EV_DONT:
   The WONT and DONT events are sent when the remote end of the
   connection wishes to disable an option, when they are refusing to
   a support an option that you have asked for, or in confirmation of
   an option you have asked to be disabled.

   Most commonly WONT and DONT events are sent as rejections of
   features you requested by sending DO or WILL events.  Receiving
   these events means the TELNET option is not or will not be
   supported by the remote end, so give up.

   Sometimes WONT or DONT will be sent for TELNET options that are
   already enabled, but the remote end wishes to stop using.  You
   cannot decline.  These events are demands that must be complied
   with.  libtelnet will always send the appropriate response back
   without consulting your application.  These events are sent to
   allow your application to disable its own use of the features.

   In either case, the TELNET option under negotiation will be in
   event->telopt field.

   Note that in PROXY mode libtelnet will do no processing of its
   own for you.

 TELNET_EV_SUBNEGOTIATION:
   Triggered whenever a TELNET sub-negotiation has been received.
   Sub-negotiations include the NAWS option for communicating
   terminal size to a server, the NEW-ENVIRON and TTYPE options for
   negotiating terminal features, and MUD-centric protocols such as
   ZMP, MSSP, and MCCP2.

   The event->telopt value is the option under sub-negotiation.  The
   remaining data (if any) is passed in event->buffer and
   event->size.  Note that most subnegotiation commands can include
   embedded NUL bytes in the subnegotiation data, and the data
   event->buffer is not NUL terminated, so always use the event->size
   value!

   The meaning and necessary processing for subnegotiations are
   defined in various TELNET RFCs and other informal specifications.
   A subnegotiation should never be sent unless the specific option
   has been enabled through the use of the telnet negotiation
   feature.

 TELNET_EV_COMPRESS
   The COMPRESS event notifies the app that COMPRESS2/MCCP2
   compression has begun or ended.  Only servers can send compressed
   data, and hence only clients will receive compressed data.

   The event->command value will be 1 if compression has started and
   will be 0 if compression has ended.
 
 TELNET_EV_WARNING
   The WARNING event is sent whenever something has gone wrong inside
   of libtelnet (possibly due to malformed data sent by the other
   end) but which recovery is (likely) possible.  It may be safe to
   continue using the connection, but some data may have been lost or
   incorrectly interpreted.

   The event->buffer value will contain a NUL terminated string
   explaining the error, and the event->size value containers the
   length of the string.

 TELNET_EV_ERROR
   Similar to the WARNING event, the ERROR event is sent whenever
   something has gone wrong.  ERROR events are non-recoverable,
   however, and the application should immediately close the
   connection.  Whatever has happened is likely going only to result
   in garbage from libtelnet.  This is most likely to happen when a
   COMPRESS2 stream fails, but other problems can occur.

   The event->buffer value will contain a NUL terminated string
   explaining the error, and the event->size value containers the
   length of the string.

III. INTEGRATING LIBTELNET WITH COMMON MUDS
=====================================================================

FIXME: fill in some notes about how to splice in libtelnet with
common Diku/Merc/Circle/etc. MUD codebases.

IV. SAFETY AND CORRECTNESS CONSIDERATIONS
=====================================================================

Your existing application may make heavy use of its own output
buffering and transmission commands, including hand-made routines for
sending TELNET commands and sub-negotiation requests.  There are at
times subtle issues that need to be handled when communication over
the TELNET protocol, not least of which is the need to escape any
byte value 0xFF with a special TELNET command.

For these reasons, it is very important that applications making use
of libtelnet always make use of the telnet_send_*() family of
functions for all data being sent over the TELNET connection.

In particular, if you are writing a client, all user input must be
passed through to telnet_send_data().  This also includes any input
generated automatically by scripts, triggers, or macros.

For a server, any and all output -- including ANSI/VT100 escape
codes, regular text, newlines, and so on -- must be passed through to
telnet_send_data().

Any TELNET commands that are to be sent must be given to one of the
following: telnet_send_command, telnet_send_negotiate, or
telnet_send_subnegotiation().

If you are attempting to enable COMPRESS2/MCCP2, you must use the
telnet_begin_compress2() function.

V. MCCP2 COMPRESSION
=====================================================================

The MCCP2 (COMPRESS2) TELNET extension allows for the compression of
all traffic sent from server to client.  For more information:

 http://www.mudbytes.net/index.php?a=articles&s=mccp

In order for libtelnet to support MCCP2, zlib must be installed and
enabled when compiling libtelnet.  Use -DHAVE_ZLIB to enable zlib
when compiling libtelnet.c and pass -lz to the linker to link in the
zlib shared library.

libtelnet transparently supports MCCP2.  For a server to support
MCCP2, the application must begin negotiation of the COMPRESS2 option
using telnet_send_negotiate(), for example:

 telnet_send_negotiate(&telnet, TELNET_WILL,
     TELNET_OPTION_COMPRESS2, user_data);

If a favorable DO COMPRESS2 is sent back from the client then the
server application can begin compression at any time by calling
telnet_begin_compress2().

If a connection is in PROXY mode and COMPRESS2 support is enabled
then libtelnet will automatically detect the start of a COMPRESS2
stream, in either the sending or receiving direction.

VI. TELNET PROXY UTILITY
=====================================================================

The telnet-proxy utility is a small application that serves both as a
testbed for libtelnet and as a powerful debugging tool for TELNET
servers and clients.

To use telnet-proxy, you must first compile it using:

 $ make

If you do not have zlib installed and wish to disable MCCP2 support
then you must first edit the Makefile and remove the -DHAVE_ZLIB and
the -lz from the compile flags.

To run telnet-proxy, you simply give it the server's host name or IP
address, the server's port number, and the port number that
telnet-proxy should listen on.  For example, to connect to the server
on mud.example.com port 7800 and to listen on port 5000, run:

 $ ./telnet-proxy mud.example.com 7800 5000

You can then connect to the host telnet-proxy is running on (e.g.
127.0.0.1) on port 500 and you will automatically be proxied into
mud.example.com.

telnet-proxy will display status information about the data passing
through both ends of the tunnel.  telnet-proxy can only support a
single tunnel at a time.  It will continue running until an error
occurs or a terminating signal is sent to the proxy process.