Distributed programming: Difference between revisions

Content added Content deleted

Inline

Revision as of 04:44, 2 July 2011

Write two programs (or one program with two modes) which run on networked computers, and send some messages between them.

The protocol used may be language-specific or not, and should be suitable for general distributed programming; that is, the protocol should be generic (not designed just for the particular example application), readily capable of handling the independent communications of many different components of a single application, and the transferring of arbitrary data structures natural for the language.

This task is intended to demonstrate high-level communication facilities beyond just creating sockets.

Ada

Works with: GNAT GPL version 2010

Works with: PolyORB

Ada defines facilities for distributed systems in its standard (Annex E, also called DSA).

This example works with PolyORB and the GNAT GPL 2010 compiler from AdaCore.

server.ads: <lang Ada>package Server is

  pragma Remote_Call_Interface;
  procedure Foo;
  function Bar return Natural;

end Server;</lang>

server.adb: <lang Ada>package body Server is

  Count : Natural := 0;

  procedure Foo is
  begin
     Count := Count + 1;
  end Foo;

  function Bar return Natural is
  begin
     return Count;
  end Bar;

end Server;</lang>

client.adb: <lang Ada>with Server; with Ada.Text_IO;

procedure Client is begin

  Ada.Text_IO.Put_Line ("Calling Foo...");
  Server.Foo;
  Ada.Text_IO.Put_Line ("Calling Bar: " & Integer'Image (Server.Bar));

end Client;</lang>

required config (dsa.cfg): <lang Ada>configuration DSA is

  pragma Starter (None);

  -- Server
  Server_Partition : Partition := (Server);
  procedure Run_Server is in Server_Partition;

  -- Client
  Client_Partition : Partition;
  for Client_Partition'Termination use Local_Termination;
  procedure Client;
  for Client_Partition'Main use Client;

end DSA;</lang>

compilation:

$po_gnatdist dsa.cfg
[...]
 ------------------------------
 ---- Configuration report ----
 ------------------------------
Configuration :
   Name        : dsa
   Main        : run_server
   Starter     : none

Partition server_partition
   Main        : run_server
   Units       :
             - server (rci)
             - run_server (normal)
             - polyorb.dsa_p.partitions (rci, from PCS)

   Environment variables :
             - "POLYORB_DSA_NAME_SERVICE"

Partition client_partition
   Main        : client
   Termination : local
   Units       :
             - client (normal)

   Environment variables :
             - "POLYORB_DSA_NAME_SERVICE"

 -------------------------------
[...]

preparation (run PolyORB name service):

$ po_ioc_naming
POLYORB_CORBA_NAME_SERVICE=IOR:010000002b00000049444[...]
POLYORB_CORBA_NAME_SERVICE=corbaloc:iiop:1.2@10.200.[...]

You have to set the environment variable POLYORB_DSA_NAME_SERVICE to one of the two values given by po_ioc_naming for the server/client partitions.

running server:

$ ./server_partition

running client:

$ ./client_partition
Calling Foo...
Calling Bar:  1
$ ./client_partition
Calling Foo...
Calling Bar:  2

AutoHotkey

See Distributed program/AutoHotkey.

C

Using PVM [[1] This program is in a sense both a server and a client, depending on if its task is spawned with a command-line argument: if yes, it spawns another task of the same executible on the parallel virtual machine and waits for it to transmit data; if no, it transmits data and is done. <lang C>#include <stdio.h>

include <stdlib.h>
include <pvm3.h>

int main(int c, char **v) { int tids[10]; int parent, spawn; int i_data, i2; double f_data;

if (c > 1) { spawn = pvm_spawn("/tmp/a.out", 0, PvmTaskDefault, 0, 1, tids); if (spawn <= 0) { printf("Can't spawn task\n"); return 1; }

printf("Spawning successful\n");

/* pvm_recv(task_id, msgtag). msgtag identifies what kind of data it is,

		 * for here: 1 = (int, double), 2 = (int, int)

* The receiving order is intentionally swapped, just to show. * task_id = -1 means "receive from any task" */ pvm_recv(-1, 2); pvm_unpackf("%d %d", &i_data, &i2); printf("got msg type 2: %d %d\n", i_data, i2);

pvm_recv(-1, 1); pvm_unpackf("%d %lf", &i_data, &f_data); printf("got msg type 1: %d %f\n", i_data, f_data); } else { parent = pvm_parent();

pvm_initsend(PvmDataDefault); i_data = rand(); f_data = (double)rand() / RAND_MAX; pvm_packf("%d %lf", i_data, f_data); pvm_send(parent, 1); /* send msg type 1 */

pvm_initsend(PvmDataDefault); i2 = rand(); pvm_packf("%d %d", i_data, i2); pvm_send(parent, 2); /* send msg type 2 */ }

pvm_exit(); return 0; }</lang>Running it: (on PVM console, exe is /tmp/a.out)<lang>pvm> spawn -> /tmp/a.out 1 spawn -> /tmp/a.out 1 [2] 1 successful t40028 pvm> [2:t40029] EOF [2:t40028] Spawning successful [2:t40028] got msg type 2: 1804289383 1681692777 [2:t40028] got msg type 1: 1804289383 0.394383 [2:t40028] EOF [2] finished</lang>

C#

This example is incorrect. Please fix the code and remove this message.

Details: The protocol used is not sufficiently general-purpose.

The example server can handle one client at any one time. It will read what the client writes, and respond with "Hello World!". The client will write "Hello World!" and read the response from the server.

Server

<lang csharp> using System.Net.Sockets;

class Program { static void Main(string[] args) { TcpListener server = new TcpListener(8000); server.Start();

Console.WriteLine("Listening, port 8000");

TcpClient client; do { // Accept client client = server.AcceptTcpClient(); Console.WriteLine("Recieved client: " + client.Client.AddressFamily.ToString());

// Recieve string tRecieve = ""; char t; do { if (client.Available > 0) { t = (char)client.GetStream().ReadByte();

if (t == 0) break;

tRecieve += t; } } while (true);

Console.WriteLine("Recieved: " + tRecieve);

// Send byte[] tSend = Encoding.ASCII.GetBytes("Hello World!"); client.GetStream().Write(tSend, 0, tSend.Length); client.GetStream().WriteByte(0);

Console.WriteLine("Sent: " + Encoding.ASCII.GetString(tSend));

// Close client.Close(); } while (true);

} } </lang>

Client

<lang csharp> using System.Net.Sockets;

class Program { static void Main(string[] args) { TcpClient client;

// Connect do { client = new TcpClient(); client.Connect(new System.Net.IPEndPoint(System.Net.IPAddress.Parse("127.0.0.1"), 8000));

Console.WriteLine("Connected");

// Send byte[] tSend = Encoding.ASCII.GetBytes("Hello World!"); client.GetStream().Write(tSend, 0, tSend.Length); client.GetStream().WriteByte(0);

Console.WriteLine("Sent: " + Encoding.ASCII.GetString(tSend));

// Read string tRecieve = ""; char t; do { if (client.Available > 0) { t = (char)client.GetStream().ReadByte();

if (t == 0) break;

tRecieve += t; } } while (true);

Console.WriteLine("Recieved: " + tRecieve);

client.Close();

Console.Read(); } while (true); } } </lang>

D

This example is incorrect. Please fix the code and remove this message.

Details: The protocol used is not sufficiently general-purpose.

Works with: Tango

Server

<lang D>module distributedserver ; import tango.net.ServerSocket, tango.text.convert.Integer,

      tango.text.Util, tango.io.Stdout ;

void main() {

 auto Ip = new InternetAddress("localhost", 12345) ;    
 auto server = new ServerSocket(Ip) ;
 auto socket = server.accept ;     
 auto buffer = new char[socket.bufferSize] ;

 bool quit = false ;
 
 while(!quit) {
   bool error = false ;    
   
   try {
     auto len = socket.input.read(buffer) ;
     auto cmd = (len > 0) ? delimit(buffer[0..len], " ") : [""] ;              
     Stdout(cmd).newline.flush ;
     switch (cmd[0]) {
       case "square":
         socket.output.write(toString(toInt(cmd[1]) * toInt(cmd[1]))) ; break ;
       case"add":
         socket.output.write(toString(toInt(cmd[1]) + toInt(cmd[2]))) ; break ;
       case "quit": 
         socket.output.write("Server Shut down") ;           
         quit = true ; break ;
       default: error = true ;
     }
   } catch (Exception e) 
     error = true ;
   if(error) socket.output.write("<Error>") ;    
   if(socket) socket.close ;
   if(!quit) socket = server.accept ;     
 }

 if(socket) socket.close ;

}</lang>

Client

<lang D>module distributedclient ; import tango.net.SocketConduit, tango.net.InternetAddress,

      tango.text.Util, tango.io.Stdout ;

void main(char[][] args) {

 if(args.length> 1) {
   try {
     auto Ip = new InternetAddress("localhost", 12345) ;    
     auto socket = new SocketConduit ;     
     socket.connect(Ip) ;
     auto buffer = new char[socket.bufferSize] ;
 
     socket.output.write(join(args[1..$]," ")) ;
     auto len = socket.input.read(buffer) ;    
     if(len > 0) Stdout(buffer[0..len]).newline ;
   
     if(socket) socket.close ;
   } catch(Exception e) 
     Stdout(e.msg).newline ;
 } else
   Stdout("usage: supply argument as,\n\tquit\n"
     "\tsquare <number>\n\tadd <number> <number>").newline ;

}</lang>

E

Protocol: Pluribus

This service cannot be used except by clients which know the URL designating it, messages are encrypted, and the client authenticates the server. However, it is vulnerable to denial-of-service by any client knowing the URL.

Server

(The protocol is symmetric; this program is the server only in that it is the one which is started first and exports an object.)

<lang E>def storage := [].diverge()

def logService {

 to log(line :String) {
   storage.push([timer.now(), line])
 }
 to search(substring :String) {
   var matches := []
   for [time, line] ? (line.startOf(substring) != -1) in storage {
     matches with= [time, line]
   }
   return matches
 }

}

introducer.onTheAir() def sturdyRef := makeSturdyRef.temp(logService) println(<captp>.sturdyToURI(sturdyRef)) interp.blockAtTop()</lang>

This will print the URL of the service and run it until aborted.

Client

The URL provided by the server is given as the argument to this program.

<lang E>def [uri] := interp.getArgs() introducer.onTheAir() def sturdyRef := <captp>.sturdyFromURI(uri) def logService := sturdyRef.getRcvr()

logService <- log("foot") logService <- log("shoe")

println("Searching...") when (def result := logService <- search("foo")) -> {

 for [time, line] in result {
   println(`At $time: $line`)
 }

}</lang>

Erlang

The protocol is erlang's own

Server

srv.erl

<lang erlang>-module(srv). -export([start/0, wait/0]).

start() ->

  net_kernel:start([srv,shortnames]),
  erlang:set_cookie(node(), rosetta),
  Pid = spawn(srv,wait,[]),
  register(srv,Pid),
  io:fwrite("~p ready~n",[node(Pid)]),
  ok.

wait() ->

  receive
      {echo, Pid, Any} ->
          io:fwrite("-> ~p from ~p~n", [Any, node(Pid)]),
          Pid ! {hello, Any},
          wait();
      Any -> io:fwrite("Error ~p~n", [Any])
  end.</lang>

Client

client.erl

<lang erlang>-module(client). -export([start/0, wait/0]).

start() ->

  net_kernel:start([client,shortnames]),
  erlang:set_cookie(node(), rosetta),
  {ok,Srv} = init:get_argument(server),
  io:fwrite("conencting to ~p~n", [Srv]),
  {srv, list_to_atom(Srv)} ! {echo,self(), hi},
  wait(),
  ok.

wait() ->

  receive
      {hello, Any} -> io:fwrite("Received ~p~n", [Any]);
      Any -> io:fwrite("Error ~p~n", [Any])
  end.</lang>

running it (*comes later)

|erlc srv.erl
|erl -run srv start -noshell
 srv@agneyam ready
*-> hi from client@agneyam

|erlc client.erl
|erl -run client start -run init stop -noshell -server srv@agneyam
 conencting to "srv@agneyam"
 Received hi

Go

Shown here is netchan, a standard Go library that enables Go channel operations across network connections. The significance is that these are type-safe data transfers of native Go types. Channels can be of any Go type although only an int channel is shown here. Netchans allow for arbitrary connections between computers, client and server roles are not mandatory. The netchan interface is independent of the type of network connection, TCP is used here.

Exporter

<lang go>package main

import (

   "fmt"
   "net"
   "netchan"

)

func main() {

   // channels to be exported, created as usual
   squareCh := make(chan int)
   resultCh := make(chan int)

   // create exporter for the two channels
   exp := netchan.NewExporter()
   err := exp.Export("square", squareCh, netchan.Recv)
   if err != nil {
       fmt.Println(err)
       return
   }
   err = exp.Export("result", resultCh, netchan.Send)
   if err != nil {
       fmt.Println(err)
       return
   }

   // create a net connection on which to publish
   listener, err := net.Listen("tcp", "127.0.0.1:0")
   if err != nil {
       fmt.Println(err)
       return
   }
   ta, _ := net.ResolveTCPAddr("tcp", listener.Addr().String())
   fmt.Println("square, result on port:", ta.Port)

   // publish channels
   go exp.Serve(listener)
   fmt.Println("Waiting for importer...")

   // use channels as usual.  here, just process a single transaction.
   n := <-squareCh
   resultCh <- n * n

   // wait for communication to complete before allowing program to terminate
   err = exp.Drain(1e8)
   if err != nil {
       fmt.Println(err)
   }

}</lang>

Importer

<lang go>package main

import (

   "fmt"
   "net"
   "netchan"
   "os"

)

func main() {

   if len(os.Args) != 2 {
       fmt.Println("usage: imp <port>")
       return
   }

   // make network connection to exporter
   conn, err := net.Dial("tcp", "127.0.0.1:"+os.Args[1])
   if err != nil {
       fmt.Println(err)
       return
   }

   // create channel importer
   imp := netchan.NewImporter(conn)

   // create channels of identical type as created in exporter process.
   squareCh := make(chan int)
   resultCh := make(chan int)

   // import connects channels in this process to matching exported
   // channels in exporter process.
   err = imp.Import("square", squareCh, netchan.Send, 1)
   if err != nil {
       fmt.Println(err)
       return
   }
   err = imp.Import("result", resultCh, netchan.Recv, 1)
   if err != nil {
       fmt.Println(err)
       return
   }

   // now use channels as usual
   squareCh <- 12
   fmt.Println("12 squared is", <-resultCh)

}</lang> Exporter is started first. Output:

square, result on port: 51951
Waiting for importer...

Importer session:

> imp 51951
12 squared is 144

JavaScript

Works with: node.js

Server

<lang javascript>var net = require('net')

var server = net.createServer(function (c){

 c.write('hello\r\n')
 c.pipe(c) // echo messages back

})

server.listen(3000, 'localhost') </lang>

Client

<lang javascript>var net = require('net')

conn = net.createConnection(3000, '192.168.1.x')

conn.on('connect', function(){ console.log('connected') conn.write('test') })

conn.on('data', function(msg){ console.log(msg.toString()) })</lang>

Objective-C

Distributed Objects are natural to Objective-C, and OpenStep and derivated framework offers an easy way of using remote objects as if it were local. The client must only know the protocol the remote object support. For the rest, calling a remote object's method or local object's method is transparent.

Works with: GNUstep

Server

The server vending the object with the name DistributedAction

ActionObjectProtocol.h <lang objc>#import <Foundation/Foundation.h> // our protocol allows "sending" "strings", but we can implement // everything we could for a "local" object @protocol ActionObjectProtocol - (NSString *)sendMessage: (NSString *)msg; @end</lang>

ActionObject.h <lang objc>#import <Foundation/Foundation.h>

import "ActionObjectProtocol.h"

@interface ActionObject : NSObject <ActionObjectProtocol>

 // we do not have much for this example!

@end</lang>

ActionObject.m <lang objc>#import <Foundation/Foundation.h>

import "ActionObject.h"

@implementation ActionObject -(NSString *)sendMessage: (NSString *)msg {

 NSLog(@"client sending message %@", msg);
 return @"server answers ...";

} @end</lang>

server.m <lang objc>#import <Foundation/Foundation.h>

import "ActionObject.h"

int main (void) {

 NSAutoreleasePool *pool;
 ActionObject *action;
 NSConnection *connect;
 NSSocketPort *port;
 
 pool = [[NSAutoreleasePool alloc] init];
 
 action = [[ActionObject alloc] init];

 port = (NSSocketPort *)[NSSocketPort port];
 // initWithTCPPort: 1234 and other methods are not supported yet
 // by GNUstep
 connect = [NSConnection

connectionWithReceivePort: port sendPort: port]; // or sendPort: nil

 [connect setRootObject: action];

 /* "vend" the object ActionObject as DistributedAction; on GNUstep
    the Name Server that allows the resolution of the registered name
    is bound to port 538 */
 if ([connect registerName:@"DistributedAction"

withNameServer: [NSSocketPortNameServer sharedInstance] ] == NO)

 {
   NSLog(@"can't register the server DistributedAction");
   exit(EXIT_FAILURE);
 }
 
 NSLog(@"waiting for messages...");

 [[NSRunLoop currentRunLoop] run];

 [pool release];
 return 0;

}</lang>

Client

client.m <lang objc>#import <Foundation/Foundation.h>

import "ActionObjectProtocol.h"

int main(void) {

 NSAutoreleasePool *pool;
 NSArray *args;
 id <ActionObjectProtocol> action;
 NSString *msg, *backmsg;

 pool = [[NSAutoreleasePool alloc] init];

 action = (id <ActionObjectProtocol>)
   [NSConnection
     rootProxyForConnectionWithRegisteredName: @"DistributedAction" 
     host: @"localhost"
     usingNameServer: [NSSocketPortNameServer sharedInstance] ];

 if (action == nil)
 {
   NSLog(@"can't connect to the server");
   exit(EXIT_FAILURE);
 }
 
 args = [[NSProcessInfo processInfo] arguments];

 if ([args count] == 1)
 {
   NSLog(@"specify a message");
   exit(EXIT_FAILURE);
 }
 
 msg = [args objectAtIndex: 1];

 // "send" (call the selector "sendMessage:" of the (remote) object
 // action) the first argument's text as msg, store the message "sent
 // back" and then show it in the log
 backmsg = [action sendMessage: msg];
 NSLog("%@", backmsg);

 [pool release];
 return 0;

}</lang>

OCaml

Works with: JoCaml

Minimalistic distributed logger with synchronous channels using the join calculus on top of OCaml.

Server

<lang ocaml>open Printf

let create_logger () =

 def log(text) & logs(l) =
     printf "Logged: %s\n%!" text;
     logs((text, Unix.gettimeofday ())::l) & reply to log

  or search(text) & logs(l) =
     logs(l) & reply List.filter (fun (line, _) -> line = text) l to search
 in
   spawn logs([]);
   (log, search)

def wait() & finished() = reply to wait

let register name service = Join.Ns.register Join.Ns.here name service

let () =

 let log, search = create_logger () in
   register "log" log;
   register "search" search;
   Join.Site.listen (Unix.ADDR_INET (Join.Site.get_local_addr(), 12345));
   wait ()</lang>

Client

<lang ocaml>open Printf

let ns_there = Join.Ns.there (Unix.ADDR_INET (Join.Site.get_local_addr(), 12345))

let lookup name = Join.Ns.lookup ns_there name

let log : string -> unit = lookup "log" let search : string -> (string * float) list = lookup "search"

let find txt =

 printf "Looking for %s...\n" txt;
 List.iter (fun (line, time) ->
              printf "Found: '%s' at t = %f\n%!" (String.escaped line) time)
   (search txt)

let () =

 log "bar";
 find "foo";
 log "foo";
 log "shoe";
 find "foo"</lang>

Oz

We show a program that starts a server on a remote machine, exchanges two messages with that server and finally shuts it down.

<lang oz>declare

 functor ServerCode
 export
    port:Prt
 define
    Stream
    Prt = {NewPort ?Stream}
    thread

for Request#Reply in Stream do case Request of echo(Data) then Reply = Data [] compute(Function) then Reply = {Function} end end

    end
 end

 %% create the server on some machine
 %% (just change "localhost" to some machine
 %% that you can use with a passwordless rsh login
 %% and that has the same Mozart version installed)
 RM = {New Remote.manager init(host:localhost)}

 %% execute the code encapsulated in the ServerCode functor
 Server = {RM apply(ServerCode $)}

 %% Shortcut: send a message to Server and receive a reply
 fun {Send X}
    {Port.sendRecv Server.port X}
 end

in

 %% echo
 {System.showInfo "Echo reply: "#{Send echo(hello)}}

 %% compute
 {System.showInfo "Result of computation: "#
  {Send compute(fun {$} 8 div 4 end)}}

 %% shut down server
 {RM close}</lang>

Perl

Using Data::Dumper and Safe to transmit arbitrary data structures as serialized text between hosts. Same code works as both sender and receiver. <lang Perl>use Data::Dumper; use IO::Socket::INET; use Safe;

sub get_data { my $sock = new IO::Socket::INET LocalHost => "localhost", LocalPort => "10000", Proto => "tcp", Listen => 1, Reuse => 1; unless ($sock) { die "Socket creation failure" } my $cli = $sock->accept();

# of course someone may be tempted to send you 'system("rm -rf /")', # to be safe(r), use Safe:: my $safe = new Safe; my $x = $safe->reval(join("", <$cli>)); close $cli; close $sock; return $x; }

sub send_data { my $host = shift; my $data = shift; my $sock = new IO::Socket::INET PeerAddr => "$host:10000", Proto => "tcp", Reuse => 1;

unless ($sock) { die "Socket creation failure" }

print $sock Data::Dumper->Dump([$data]); close $sock; }

if (@ARGV) { my $x = get_data(); print "Got data\n", Data::Dumper->Dump([$x]); } else { send_data('some_host', { a=>100, b=>[1 .. 10] }); }</lang>

PicoLisp

Server

<lang PicoLisp>(task (port 12321) # Background server task

  (let? Sock (accept @)
     (unless (fork)                   # Handle request in child process
        (in Sock
           (while (rd)                # Handle requests
              (out Sock
                 (pr (eval @)) ) ) )  # Evaluate and send reply
        (bye) )                       # Exit child process
     (close Sock) ) )                 # Close socket in parent process</lang>

Client

<lang PicoLisp>(let? Sock (connect "localhost" 12321)

  (out Sock (pr '*Pid))               # Query PID from server
  (println 'PID (in Sock (rd)))       # Receive and print reply
  (out Sock (pr '(* 3 4)))            # Request some calculation
  (println 'Result (in Sock (rd)))    # Print result
  (close Sock) )                      # Close connection to server</lang>

Output:

PID 18372
Result 12

Python

Works with: Python version 2.4 and 2.6

XML-RPC

Protocol: XML-RPC

Server

<lang python>#!/usr/bin/env python

-*- coding: utf-8 -*-

import SimpleXMLRPCServer

class MyHandlerInstance:

   def echo(self, data):
       Method for returning data got from client
       return 'Server responded: %s' % data

   def div(self, num1, num2):
       Method for divide 2 numbers
       return num1/num2

def foo_function():

   A function (not an instance method)
   return True

HOST = "localhost" PORT = 8000

server = SimpleXMLRPCServer.SimpleXMLRPCServer((HOST, PORT))

register built-in system.* functions.

server.register_introspection_functions()

register our instance

server.register_instance(MyHandlerInstance())

register our function as well

server.register_function(foo_function)

try:

   # serve forever
   server.serve_forever()

except KeyboardInterrupt:

   print 'Exiting...'
   server.server_close()</lang>

Client

<lang python>#!/usr/bin/env python

-*- coding: utf-8 -*-

import xmlrpclib

HOST = "localhost" PORT = 8000

rpc = xmlrpclib.ServerProxy("http://%s:%d" % (HOST, PORT))

print what functions does server support

print 'Server supports these functions:', print ' '.join(rpc.system.listMethods())

echo something

rpc.echo("We sent this data to server")

div numbers

print 'Server says: 8 / 4 is: %d' % rpc.div(8, 4)

control if foo_function returns True

if rpc.foo_function():

   print 'Server says: foo_function returned True'</lang>

HTTP

Protocol: HTTP

Server

<lang python>#!/usr/bin/python

-*- coding: utf-8 -*-

import BaseHTTPServer

HOST = "localhost" PORT = 8000

we just want to write own class, we replace do_GET method. This could be extended, I just added basics
see; http://docs.python.org/lib/module-BaseHTTPServer.html

class MyHTTPHandler(BaseHTTPServer.BaseHTTPRequestHandler):

   def do_GET(self):
       # send 200 (OK) message
       self.send_response(200)
       # send header
       self.send_header("Content-type", "text/html")
       self.end_headers()

       # send context
       self.wfile.write("<html><head><title>Our Web Title</title></head>")

self.wfile.write("<body>

This is our body. You wanted to visit %s page

</body>" % self.path)

       self.wfile.write("</html>")

if __name__ == '__main__':

   server = BaseHTTPServer.HTTPServer((HOST, PORT), MyHTTPHandler)
   try:
       server.serve_forever()
   except KeyboardInterrupt:
       print 'Exiting...'
       server.server_close()</lang>

Client

<lang python>#!/usr/bin/python

-*- coding: utf-8 -*-

import httplib

HOST = "localhost" PORT = 8000

conn = httplib.HTTPConnection(HOST, PORT) conn.request("GET", "/somefile")

response = conn.getresponse() print 'Server Status: %d' % response.status

print 'Server Message: %s' % response.read()</lang>

Socket, Plain Text

This example is incomplete. Please ensure that it meets all task requirements and remove this message.

Protocol: Plain Text

Use with Pythons pickle module for data serialization into printable text would allow the transfer of arbitrary Python data, but as it stands, this method is too low level to fulfill the task.

Server

<lang python>#!/usr/bin/python

-*- coding: utf-8 -*-

import SocketServer

HOST = "localhost" PORT = 8000

our instance that will upper whatever it gets and send back to client

class UpperCaseHandler(SocketServer.StreamRequestHandler):

   def handle(self):
       print '%s connected' % self.client_address[0]
       # get what client sends
       get = self.rfile.readline()
       # write back to client
       self.wfile.write(get.upper())

if __name__ == '__main__':

   tcpserver = SocketServer.TCPServer((HOST, PORT), UpperCaseHandler)
   try:
       tcpserver.serve_forever()
   except KeyboardInterrupt:
       print 'Exiting...'
       tcpserver.server_close()</lang>

Client

<lang python>#!/usr/bin/python

-*- coding: utf-8 -*-

import socket

HOST = "localhost" PORT = 8000

DATA = "my name is eren"

connect to server and send data

sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM) sock.connect((HOST, PORT)) sock.send("%s\n" % DATA)

get

response = sock.recv(256) sock.close()

print "We sent: %s" % DATA print 'Server responded: %s' % response</lang>

Pyro

Note: You should install Pyro (http://pyro.sourceforge.net) first and run pyro-ns binary to run code below.

Server

<lang python>#!/usr/bin/python

-*- coding: utf-8 -*-

import Pyro.core import Pyro.naming

create instance that will return upper case

class StringInstance(Pyro.core.ObjBase):

   def makeUpper(self, data):
       return data.upper()

class MathInstance(Pyro.core.ObjBase):

   def div(self, num1, num2):
       return num1/num2

if __name__ == '__main__':

   server = Pyro.core.Daemon()
   name_server = Pyro.naming.NameServerLocator().getNS()
   server.useNameServer(name_server)
   server.connect(StringInstance(), 'string')
   server.connect(MathInstance(), 'math')
   try:
       server.requestLoop()
   except KeyboardInterrupt:
       print 'Exiting...'
       server.shutdown()</lang>

Client

<lang python>#!/usr/bin/python

-*- coding: utf-8 -*-

import Pyro.core

DATA = "my name is eren" NUM1 = 10 NUM2 = 5

string = Pyro.core.getProxyForURI("PYRONAME://string") math = Pyro.core.getProxyForURI("PYRONAME://math")

print 'We sent: %s' % DATA print 'Server responded: %s\n' % string.makeUpper(DATA)

print 'We sent two numbers to divide: %d and %d' % (NUM1, NUM2) print 'Server responded the result: %s' % math.div(NUM1, NUM2)</lang>

Spread

Note: You should install Spread (http://www.spread.org) and its python bindings (http://www.python.org/other/spread/)

Server

You don't need any code for server. You should start "spread" daemon by typing "spread -c /etc/spread.conf -n localhost". If you want more configuration, look at /etc/spread.conf.

After starting daemon, if you want to make sure that it is running, enter spuser -s 4803 command where 4803 is your port set in spread.conf, you will see prompt, type j user, you should see something like this message: Received REGULAR membership for group test with 3 members, where I am member 2

Client (Listener)

<lang python>#!/usr/bin/python

-*- coding: utf-8 -*-

import spread

PORT = '4803'

connect spread daemon

conn = spread.connect(PORT)

join the room

conn.join('test')

print 'Waiting for messages... If you want to stop this script, please stop spread daemon' while True:

   recv = conn.receive()
   if hasattr(recv, 'sender') and hasattr(recv, 'message'):
       print 'Sender: %s' % recv.sender
       print 'Message: %s' % recv.message</lang>

Client (Sender)

<lang python>#!/usr/bin/python

-*- coding: utf-8 -*-

import spread

PORT = '4803'

conn = spread.connect(PORT) conn.join('test')

conn.multicast(spread.RELIABLE_MESS, 'test', 'hello, this is message sent from python') conn.disconnect()</lang>

Ruby

Uses

Library: dRuby

The "druby:" protocol uses TCP/IP sockets for communication.

Server <lang ruby>require 'drb/drb'

The URI for the server to connect to

URI="druby://localhost:8787"

class TimeServer

 def get_current_time
   return Time.now
 end

end

The object that handles requests on the server

FRONT_OBJECT = TimeServer.new

$SAFE = 1 # disable eval() and friends

DRb.start_service(URI, FRONT_OBJECT)

Wait for the drb server thread to finish before exiting.

DRb.thread.join</lang>

Client <lang ruby>require 'drb/drb'

The URI to connect to

SERVER_URI = "druby://localhost:8787"

Start a local DRbServer to handle callbacks.
Not necessary for this small example, but will be required
as soon as we pass a non-marshallable object as an argument
to a dRuby call.

DRb.start_service

timeserver = DRbObject.new_with_uri(SERVER_URI) puts timeserver.get_current_time</lang>

Tcl

A rudimentary IRC Server <lang tcl>proc main {} {

   global connections
   set connections [dict create]
   socket -server handleConnection 12345
   vwait dummyVar ;# enter the event loop

}

proc handleConnection {channel clientaddr clientport} {

   global connections
   dict set connections $channel address "$clientaddr:$clientport"
   fconfigure $channel -buffering line
   fileevent $channel readable [list handleMessage $channel]

}

proc handleMessage {channel} {

   global connections
   if {[gets $channel line] == -1} {
       disconnect $channel
   } else {
       if {[string index [string trimleft $line] 0] eq "/"} {
           set words [lassign [split [string trim $line]] command]
           handleCommand $command $words $channel
       } else {
           echo $line $channel
       }
   }

}

proc disconnect {channel} {

   global connections
   dict unset connections $channel
   fileevent $channel readable ""
   close $channel

}

proc handleCommand {command words channel} {

   global connections
   switch -exact -- [string tolower $command] {
       /nick {
           dict set connections $channel nick [lindex $words 0]
       }
       /quit {
           echo bye $channel
           disconnect $channel
       }
       default {
           puts $channel "\"$command\" not implemented"
       }
   }

}

proc echo {message senderchannel} {

   global connections
   foreach channel [dict keys $connections] {
       if {$channel ne $senderchannel} {
           set time [clock format [clock seconds] -format "%T"]
           set nick [dict get $connections $channel nick]
           puts $channel [format "\[%s\] %s: %s" $time $nick $message]
       }
   }

}

main</lang> Client <lang tcl>proc main {} {

   global argv argc
   if {$argc != 2} {
       error "usage: [info script] serveraddress serverport"
   }
   connect {*}$argv
   vwait dummyVar

}

proc connect {addr port} {

   global sock
   set sock [socket $addr $port]
   fconfigure $sock -buffering line
   fileevent $sock readable getFromServer
   fileevent stdin readable sendToServer

}

proc getFromServer {} {

   global sock
   if {[gets $sock line] == -1} {
       puts "disconnected..."
       exit
   } else {
       puts $line
   }

}

proc sendToServer {} {

   global sock
   set msg [string trim [gets stdin]]
   if {[string length $msg] > 0} {
       puts $sock $msg
   }

}

main</lang>

UnixPipes

This example is incorrect. Please fix the code and remove this message.

Details: Not high-level enough; does not provide support for data structures nor independent components.

Uses netcat and a buffer to cycle the server shell's stdout back to netcat's stdin

Server

<lang bash>tail -F /tmp/buffer | nc -l -p 1234 | &> /tmp/buffer</lang>

Client

Now you can enter commands in the client terminal and get the output back through the same connection.