chickadee » tcp6

tcp6

tcp6 provides facilities for communicating over TCP/IP and supports both IPv4 and IPv6.

Overview

This extension provides facilities for communicating over TCP sockets with an interface that is backwards-compatible with Unit tcp. It is implemented on top of the socket egg and consequently supports IPv4 and IPv6, non-blocking operations on Windows, and correct error detection on Windows.

All errors related to failing network operations will raise a condition of kind (exn i/o net). Timeouts raise the error (exn i/o net timeout).

Usage

Simply replace (use tcp) with (use tcp6). The API is the same, although it includes a few extensions: tcp-bind-v6-only, tcp-connect/ai and tcp-listener-socket.

Servers

(tcp-listen TCPPORT [BACKLOG [HOST]]) procedure

Creates and returns a TCP listener object that listens for connections on TCPPORT, which should be an exact integer. BACKLOG specifies the number of maximally pending connections (and defaults to 10).

If the optional argument HOST is given and not #f, then only incoming connections for the given host (or IP) are accepted.

When HOST is #f (the default), the behavior is system-dependent. It should listen on all IPv4 and IPv6 addresses if possible, or just on IPv4 if IPv6 is disabled. This is true on OS X and Windows. Unfortunately, certain systems may always prefer to listen on IPv4 only (particularly those using recent glibc, like Ubuntu).

Special note when HOST is #f. If you have set (tcp-bind-ipv6-only #t), or if tcp-bind-ipv6-only is not supported by your OS, we always listen on "0.0.0.0" to IPv4 only. This is done because users will expect (tcp-listen port) to listen at least on IPv4. To listen to IPv6 only in this case, explicitly specify a HOST of "::".

Long story short, setting HOST to #f will more likely than not give you an IPv4-only listener.

(tcp-listener? X) procedure

Returns #t if X is a TCP listener object, or #f otherwise.

(tcp-close LISTENER) procedure

Reclaims any resources associated with LISTENER.

(tcp-accept LISTENER) procedure

Waits until a connection is established on the port on which LISTENER is listening and returns two values: an input- and output-port that can be used to communicate with the remote process. The current value of tcp-accept-timeout is used to determine the maximal number of milliseconds (if any) to wait until a connection is established. When a client connects any read- and write-operations on the returned ports will use the current values (at the time of the connection) of tcp-read-timeout and tcp-write-timeout, respectively, to determine the maximal number of milliseconds to wait for input/output before a timeout error is signalled.

Note: this operation and any I/O on the ports returned will not block other running threads.

(tcp-accept-ready? LISTENER) procedure

Returns #t if there are any connections pending on LISTENER, or #f otherwise.

(tcp-listener-port LISTENER) procedure

Returns the port number assigned to LISTENER. (If you pass 0 to tcp-listen, then the system will choose a port-number for you.)

(tcp-listener-socket LISTENER) procedure

Returns the socket object associated with LISTENER. This procedure is an addition over Unit tcp.

(tcp-listener-fileno LISTENER) procedure

Returns the file-descriptor associated with LISTENER.

Clients

(tcp-connect HOSTNAME [TCPPORT]) procedure

Establishes a client-side TCP connection to the machine with the node name or IP address HOSTNAME (a string) at TCPPORT (an exact integer or a service name) and returns two values: an input- and output-port for communicating with the remote process.

If the TCPPORT is omitted, the port is parsed from the HOSTNAME string. The format expected is "HOST:PORT", or "[HOST]:PORT" if HOST is an IPv6 string. The PORT can either be a string representation of an integer or a service name which is translated to an integer using address-information from socket.

Address resolution is performed using address-information, which may return multiple addresses for a given hostname, including both IPv6 and IPv4 addresses. tcp-connect will try each of these in turn until one succeeds. See tcp-connect/ai for more information. For example, connecting to "localhost:ssh" may connect to "[::1]:22", "[fe80::1%lo0]:22", and "127.0.0.1:22" in turn, until an ssh listener is contacted.

The current value of tcp-connect-timeout is used to determine the maximum number of milliseconds (if any) to wait until the connection is established. When the connection takes place any read- and write-operations on the returned ports will use the current values (at the time of the call to tcp-connect) of tcp-read-timeout and tcp-write-timeout, respectively, to determine the maximum number of milliseconds to wait for input/output before a timeout error is signalled.

Note: any I/O on the ports returned will not block other running threads.

(tcp-connect/ai ais) procedure

Takes a list of addrinfo objects, obtained from address-information in the socket egg, and connects to each in turn until one succeeds. If a timeout occurs during connection, or a transient error (connection refused, host unreachable) occurs, the next address in the list will be tried. If all addresses fail, the last error encountered is propagated to the caller. If a fatal socket error occurs then we terminate immediately.

(tcp-connect host port) is equivalent to

(tcp-connect/ai (address-information host port))

which may include IPv6 and IPv4 addresses. To connect instead to localhost:22 over IPv4 only:

(tcp-connect/ai (address-information "localhost" 22 family: af/inet))

and to try to connect to the first HTTP mirror that accepts your connection:

(tcp-connect/ai
 (append (address-information "athena.example.com" 80)
         (address-information "achilles.example.com" 80)
         (address-information "aphrodite.example.com" 80)))

Keeping the connect timeout low is probably a good idea in the last case.

Port operations

(tcp-addresses PORT) procedure

Returns two values for the input- or output-port PORT (which should be a port returned by either tcp-accept or tcp-connect): the IP address of the local and the remote machine that are connected over the socket associated with PORT. The returned addresses are IPv4 or IPv6 strings.

(tcp-port-numbers PORT) procedure

Returns two values for the input- or output-port PORT (which should be a port returned by either tcp-accept or tcp-connect): the TCP port numbers of the local and the remote machine that are connected over the socket associated with PORT.

(tcp-abandon-port PORT) procedure

Marks the socket port PORT as abandoned. This is mainly useful to close down an input or output port without shutting down that side of the connection. See socket-abandon-port in socket for more information.

(tcp-port->socket PORT) procedure

Return the socket object associated with TCP input or output port PORT.

It is also possible to use port->fileno from Unit posix with TCP ports created by this egg.

Parameters

tcp-buffer-size parameter

Sets the size of the output buffer. By default no output-buffering for TCP output is done, but to improve performance by minimizing the number of TCP packets, buffering may be turned on by setting this parameter to an exact integer greater than zero. For best performance, it should be a power of 2 such as 128, 1024 or 4096. A buffer size of #f turns buffering off. The setting of this parameter takes effect at the time when the I/O ports for a particular socket are created, i.e. when tcp-connect or tcp-accept is called.

See socket-receive-buffer-size and socket-send-size in the socket egg for additional send and receive tuning that can be done with TCP ports. This parameter is itself equivalent to socket-send-buffer-size.

Note that since output is not immediately written to the associated socket, you may need to call flush-output once you want the output to be transmitted. Closing the output port will flush automatically.

tcp-read-timeout parameter

Determines the timeout for TCP read operations in milliseconds. A timeout of #f disables timeout checking. The default read timeout is 60000, i.e. 1 minute. If timeout occurs while reading, a condition object of kind (exn i/o net timeout) is thrown.

tcp-write-timeout parameter

Determines the timeout for TCP write operations in milliseconds. A timeout of #f disables timeout checking. The default write timeout is 60000, i.e. 1 minute. If timeout occurs while writing, a condition object of kind (exn i/o net timeout) is thrown.

tcp-connect-timeout parameter

Determines the timeout for tcp-connect operations in milliseconds. A timeout of #f disables timeout checking and is the default. If timeout occurs while trying to connect, a condition object of kind (exn i/o net timeout) is thrown.

tcp-accept-timeout parameter

Determines the timeout for tcp-accept operations in milliseconds. A timeout of #f disables timeout checking and is the default. If timeout occurs while waiting for connections, a condition object of kind (exn i/o net timeout) is thrown.

tcp-bind-ipv6-only parameter

When #f, the default, IPv6 listening sockets will accept IPv6 or IPv4 connections when possible. This is only relevant when listening on the unspecified address "::". When #t, IPv6 listeners will not accept IPv4 connections.

This option is ignored when unsupported, such as on Windows 2000 and XP, whose IPv4 and IPv6 stacks are separate.

Example

A very simple example follows. Say we have the two files client.scm and server.scm:

 ; client.scm
 (use tcp6)
 (define-values (i o) (tcp-connect "localhost" 4242))
 (write-line "Good Bye!" o)
 (print (read-line i))
 ; server.scm
 (use tcp6)
 (define l (tcp-listen 4242))
 (define-values (i o) (tcp-accept l))
 (write-line "Hello!" o)
 (print (read-line i))
 (close-input-port i)
 (close-output-port o)
% csc server.scm
% csc client.scm
% ./server &
% ./client
Good Bye!
Hello!

About this egg

Author

Jim Ursetto

This code has been completely rewritten from the core TCP unit, but some of the original code was moved into the socket egg, and the API is largely unchanged.

Version history

0.1.1
Bugfix for #963: tcp-accept now propagates read/write timeout to socket egg.
0.1
Initial release.

License

Copyright (c) 2011, Jim Ursetto
Copyright (c) 2008-2011, The Chicken Team
Copyright (c) 2000-2007, Felix L. Winkelmann
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:

- Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
- Neither the name of the author nor the names of its contributors
may be used to endorse or promote products derived from this
software without specific prior written permission.
 
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
OF THE POSSIBILITY OF SUCH DAMAGE.

Contents »