从Graphic PU走向General PU

G300 作为 nVidia 的下一代GPU，3引入很多变化: 512 CUDA Cores, L1/L2 cache, GigaThread, ECC memory controller. 看起来都更像是一个General Processing Unit.

nVidia 还是有雄心壮志的，对于nVidia而言，这是一个大胆的尝试。从架构上看，确实可以被称为下一代。

in reference to: Next Generation CUDA Architecture, Code Named Fermi (view on Google Sidewiki)

Google Voice and bandwidth.com

It isn't a secret that the land line provided by Google Voice is ordered from Bandwidth.com -- The 4st fastest growing privately held company in U.S.A. Bandwidth.com is a company in RTP, North Carolina.

The small high-tech company always be the source of creatures! Why we don't have these in China? //sigh

VIM Wiki

非常不错的网站 介绍了很多VIM的技巧。今天用到了一个小技巧，如何在VIM内排序。在VIM7.0
这个非常简单，可以直接使用内置的`:sort' 命令。当然，还可以剔除重复的 `:sort u'.

What's Byte Order Mark (BOM) in Unicode

When a BOM is used, is it only in 16-bit Unicode text?

A: No, a BOM can be used as a signature no matter how the Unicode text is transformed: UTF-16, UTF-8, UTF-7, etc. The exact bytes comprising the BOM will be whatever the Unicode character FEFF is converted into by that transformation format. In that form, the BOM serves to indicate both that it is a Unicode file, and which of the formats it is in. Examples:

Bytes	Encoding Form
00 00 FE FF	UTF-32, big-endian
FF FE 00 00	UTF-32, little-endian
FE FF	UTF-16, big-endian
FF FE	UTF-16, little-endian
EF BB BF	UTF-8

链接是 UTF系列问题的权威解答FAQ，强烈建议阅读！

VIM unicode

在 MS Windows 下面由于受 Byte Order Mark (BOM) 的控制， 需要对GVIM 做一些配置才能正常工作。 这包括 tenc, enc, fencs, guifont, guifontwide。 这样一来，vim就可以自动 根据BOM做出正确的显示了，同时也可以自动识别GBK和BIG5。

if has("multi_byte")
  set encoding=unicode
  let &termencoding = &encoding
  " Set default encoding as UTF-8 with BOM
  setglobal fileencoding=utf-8 bomb
  " Auto detect the file encoding BOM unicode, utf-8, GBK, Big5, Latin1
  set fileencodings=ucs-bom,utf-8,cp936,cp950,latin1

  " Walkaround the encoding problem for fenc=utf-8 && enc=ucs-2
  augroup i18n
      autocmd!
      autocmd BufReadPost * if &fileencoding == "utf-8" | let &encoding = &fileencoding | endif
  augroup END
endif

" Enable the Chinese characters
"set guifont=Bitstream\ Vera\ Sans\ Mono:h10:cANSI,NSimSun:h11:cGB2312
set guifont=Consolas:h11:cANSI,NSimSun:h11:cGB2312
set guifontwide=NSimSun:h11:cGB2312

Google Wave Sandbox

This is the first time for me to use Google Wave. It looks a little strange as a fresh user. We MUST fully understand "what is a WAVE?"

• IM?
  
• Word?
  
• Blog?
• twitter?
  
• SNS?
  

Are you understanding this? It might be better if there are some others in my Wave. I need to take a look at the Google Wave API to find more.

It just likes someone said, "It looks more like a Microsoft product, not Google!" I think, it still takes time to figure out "What's a Wave?"

Hex string to ASCII string in Perl

If you want to convert a string "ABCDEFG" to Hex mode like "07,41,42,43,44,45,46,47" -- a length prefix with all the others in HEX mode. How to do it in Perl?

You can do

$to = join(",", map { unpack('H2', $_) } chr(length($from)), split(/(?=.)/, $from));

or

$to = join(",", map { unpack('H2', $_) } split(/(?=.)/, pack('W/a', $from)));

How to do the reverse??? A question for you :P

An English-to-Chinese Dictionary based on the evil Google Suggest Quries

<SearchPlugin xmlns="http://www.mozilla.org/2006/browser/search/">
<ShortName>gDictionary</ShortName>
<Description>Google English-Chinese Dictionary</Description>
<InputEncoding>UTF-8</InputEncoding>
<Image width="16" height="16">data:image/png;base64,AAABAAEAEBAAAAEAGABoAwAAFgAAACgAAAAQAAAAIAAAAAEAGAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAADs9Pt8xetPtu9FsfFNtu%2BTzvb2%2B%2Fne4dFJeBw0egA%2FfAJAfAA8ewBBegAAAAD%2B%2FPtft98Mp%2BwWsfAVsvEbs%2FQeqvF8xO7%2F%2F%2F63yqkxdgM7gwE%2FggM%2BfQA%2BegBDeQDe7PIbotgQufcMufEPtfIPsvAbs%2FQvq%2Bfz%2Bf%2F%2B%2B%2FZKhR05hgBBhQI8hgBAgAI9ewD0%2B%2Fg3pswAtO8Cxf4Kw%2FsJvvYAqupKsNv%2B%2Fv7%2F%2FP5VkSU0iQA7jQA9hgBDgQU%2BfQH%2F%2Ff%2FQ6fM4sM4KsN8AteMCruIqqdbZ7PH8%2Fv%2Fg6Nc%2Fhg05kAA8jAM9iQI%2BhQA%2BgQDQu6b97uv%2F%2F%2F7V8Pqw3eiWz97q8%2Ff%2F%2F%2F%2F7%2FPptpkkqjQE4kwA7kAA5iwI8iAA8hQCOSSKdXjiyflbAkG7u2s%2F%2B%2F%2F39%2F%2F7r8utrqEYtjQE8lgA7kwA7kwA9jwA9igA9hACiWSekVRyeSgiYSBHx6N%2F%2B%2Fv7k7OFRmiYtlAA5lwI7lwI4lAA7kgI9jwE9iwI4iQCoVhWcTxCmb0K%2BooT8%2Fv%2F7%2F%2F%2FJ2r8fdwI1mwA3mQA3mgA8lAE8lAE4jwA9iwE%2BhwGfXifWvqz%2B%2Ff%2F58u%2Fev6Dt4tr%2B%2F%2F2ZuIUsggA7mgM6mAM3lgA5lgA6kQE%2FkwBChwHt4dv%2F%2F%2F728ei1bCi7VAC5XQ7kz7n%2F%2F%2F6bsZkgcB03lQA9lgM7kwA2iQktZToPK4r9%2F%2F%2F9%2F%2F%2FSqYK5UwDKZAS9WALIkFn%2B%2F%2F3%2F%2BP8oKccGGcIRJrERILYFEMwAAuEAAdX%2F%2Ff7%2F%2FP%2B%2BfDvGXQLIZgLEWgLOjlf7%2F%2F%2F%2F%2F%2F9QU90EAPQAAf8DAP0AAfMAAOUDAtr%2F%2F%2F%2F7%2B%2Fu2bCTIYwDPZgDBWQDSr4P%2F%2Fv%2F%2F%2FP5GRuABAPkAA%2FwBAfkDAPAAAesAAN%2F%2F%2B%2Fz%2F%2F%2F64g1C5VwDMYwK8Yg7y5tz8%2Fv%2FV1PYKDOcAAP0DAf4AAf0AAfYEAOwAAuAAAAD%2F%2FPvi28ymXyChTATRrIb8%2F%2F3v8fk6P8MAAdUCAvoAAP0CAP0AAfYAAO4AAACAAQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAAQAA</Image>
<Url type="application/x-suggestions+json" method="GET" template="http://suggestqueries.google.com/complete/search?client=suggest&amp;hjson=t&amp;ds=d&amp;hl=zh-CN&amp;q={searchTerms}"></Url>
<Url type="text/html" method="GET" template="http://www.google.com/dictionary?aq=f&amp;langpair=en|zh-CN&amp;hl=en&amp;q={searchTerms}"></Url>
<SearchForm>http://www.google.com/dictionary</SearchForm>
</SearchPlugin>

Small Skills of Command Prompt for Windows

There are some useful skills for CMD.exe in Windows.

1, How to do a substring in the CMD.exe -- %VAR_NAME:~POS,LEN%
for example

set name=fingercat

echo %name:~6,3%

This will give you a "cat" ;)

2, How to get the input of user as a string?

On windows 2000 or later. It is simple.

set /p answer=Prompt Message Here

echo %answer%

This will give you the input stirng of user.

Windows Command Prompt (.bat) file never suppose to be a powerful shell. But Microsoft really give it more an more power until a real PowerShell came into place.

Inside the Erlang VM - focus on SMP

It discussed the current status and possible future of SMP enabled Erlang VM.
It is worth to read!

VIMRC file

source $VIMRUNTIME/vimrc_example.vim
source $VIMRUNTIME/gvimrc_example.vim
set nobackup
set ch=1
set tabstop=4
set shiftwidth=4
"set softtabstop=4
set expandtab
colorscheme darkblue
"colorscheme oceandeep


" Enable the Chinese characters
set guifont=Bitstream\ Vera\ Sans\ Mono:h10:cANSI,NSimSun:h12:cGB2312
set guifontwide=NSimSun:h12:cGB2312


" display the syntax name under the cursor
map <F12> :echo synIDattr(synIDtrans(synID(line("."), col("."), 1)), "name")<CR>


" We can navigate the list by using the <F3> and <S-F3>
map <F3> :cnext<CR>
map <S-F3> :cprevious<CR>


" Let's help the tab navigation a little
map <F9> :tabprevious<CR>
map <S-F9> :tabnext<CR>


" Tlist
map <F5> :Tlist<CR>
set tags=./tags,../tags,../../tags

C++ reinterpret_cast

说起 C++ 的类型转换，一共有四种：

• dynamic_cast
• static_cast
• reinterpret_cast
• const_cast

其中 reinterpret_cast 是很有意思的，它可以被看作是 C 语言中强制类型转换的功能，它是一种非检查的暴力转换手段。通过它，我们可以把 指向 int* 转换为 double* 。

当然也可以用来做函数指针的转换，比如：

void FuncA(char *src, int len, int *pLen)
{
 *pLen = len;
}

int _tmain(int argc, _TCHAR* argv[])
{
 bool (*func_ptr) (char*, int);

 func_ptr = reinterpret_cast<bool (*)(char*, int)>(&FuncA);

 return 0;
}

当然了也可以用来转换 类成员函数，如：

class ObjectA
{
public:
 void FuncA(char *src, int len, int *pLen)
 {
  *pLen = len;
 }

 bool FuncB(char *src, int len, int *pLen)
 {
  *pLen = len;
  return false;
 }
};

class ObjectB
{

};

int _tmain(int argc, _TCHAR* argv[])
{
 bool (ObjectB::*func_ptr) (char*, int);

 //ObjectA obj;
 func_ptr = reinterpret_cast<bool (ObjectB::*)(char*, int)>(&ObjectA::FuncB);

 return 0;
}

但是不能在普通函数和类成员函数之间转换。

Install Safari4 without installation on Windows

First, Extract the Safari.cab into a folder. The Safari.cab is from Safari.msi of SafariSetup.exe, you can open the .exe and .msi by using 7-zip tool. :D

Second, put the file into the right folder. If you have MinGW or Cygwin, you can use this script to move all the files into the right folder.

for i in *; do fp=${i//_/\/}; echo $fp; f=$(basename $fp); eval "p=\${fp%$f}"; if [ $fp != $f ];then if [ ! -d $p ];then mkdir -p $p; fi; mv $i $fp; fi; done

Now ready to go! Safari4 beta is a green software.

NORTEL 幽默

NORTEL 从来就不缺乏有才华的人，这个我曾经工作过的百年老店今天已经是风雨飘摇。
看看这些 NORTEL 员工们制作的讽刺剧吧！

我不得不赞叹，太油菜了！

Java GUI Effectives

Java gets widely usage on Server side, while the client technology is still evolution continuously. Since Java 5.0, more and more interesting GUI feature are added into the JVM. They are:

• OpenGL-based render pipeline
  As you know, Swing renders all the widgets by itself, that's the reason why some case the Java GUI looks so strange. :) Previously on Microsoft Windows, the DirectDraw and Direct3D was the only choice. But now you can use OpenGL pipeline as well.
  It is disabled by default, but it can enabled by setting
  

  -Dsun.java2d.opengl=true
  

• Font Anti-Aliasing
  It is an important feature for the font anti-aliasing. Now it is possible for Java application to specify the Font Anti-Aliasing feature by itself with system property:
  -Dawt.useSystemAAFontSettings=lcd
  
  
• Metal and Nimbus Look and Feel
  Metal (from Java 5.0) and Nimbus (from Java 6.0u10) are available in standard JRE distribution.

Really hope Java could be better and better~~~

C++ vtable

在 C++ 的OO模型中，很重要的一个特性是运行时多态。运行时多态是通过类的虚方法实现的。看下面这个例子：

struct IObject {
  virtual void MethodA() = 0;  
  virtual int  MethodB(char *) = 0;
};

class A : public IObject {
public:
  virtual void MethodA() {
    std::cout << "method A" << std::endl;
  };

  virtual int MethodB(char *p) {
    std::cout << "method B " << p << std::endl;
    return 0;
  };

public:
  int MethodC(int c) {
    std::cout << "method C non-virtual" << std::endl;
    return ++c;
  }
private:
  int mem_;
};

int main(int argc, cahr *argv[]) 
{
  IObject *pObj = new A;

  pObj->MethodA();
  pObj->MethodB(NULL);

  // FIXME: This is a compile error!
  pObj->MethodC(1); // error

  delete A;

  return 0;
};

C++ 运行时多态， 在执行代码中，主要通过 vtable 实现。让我们用C来重新描述上面的例子。

// VTABLE of class IObject
struct I_IObject {
  void *pMethodA(void);
  int  *pMethodB(char *);
};

// Instance variant type of IObject
struct IObject {
  struct I_IObject *vtable_this;
};

// VTABLE of class A
struct A;
struct I_A {
  void *pDtor(A*) 
};

// Instance object variant type of class A
struct A {
  struct I_IObject *vtable_IObject;
  struct I_A *vtable_this;

  int *pMethodC(int);
  int mem_;
};

void A_MethodA (void) {
  printf("method A\n"); 
}

int  A_MethodB (char *p) {
  printf("method B %s\n", p);
}

int  A_MethodC (int c) {
  printf("method C non-virtual\n");
  return ++c;
}

void A_dtor(struct A *_this) {
  free(_this);
}

struct A *A_ctor() {
  struct A *p = malloc(sizeof(struct A));

  // Non-virtual function will be assign the value directly.
  p.pMethodC = &A_MethodC;

  /////////////////////////////////////////////////
  // vtable initialization code
#ifdef A_MethodA
  p.vtable_IObject->pMethodA = &A_MethodA;
#else
  p.vtable_IObject->pMethodA = &IObject_MethodA; // Warning!
#endif

#ifdef A_MethodB
  p.vtable_IObject->pMethodB = &A_MethodB;
#else
  p.vtable_IObject->pMethodB = &IObject_MethodB; // Warning!
#endif

  // vtable of A itself
  p.vtable_self->pDtor = &A_dtor;

  return p;
}

int main(int argc, char *argv[]) 
{
  IObject *pObj = (IObject*) A_ctor();

  pObj->vtable_self->pMethodA();
  pObj->vtable_self->pMethodB(NULL);

  A_dtor(pObj);

  return 0;
}

C的定价物确实有些繁琐，不过基本的结构就是这样了。具体说来是这样：
1，如果一个类 virtual function, 这些 virtual funciton 就会被统一在一个 vtable 中；
2，包含 virtual function 的class 都会有一个指向 vtable structure的指针，如果当前类重写了 virtual function, 在对象初始化的时候相应的虚函数指针就会被替换；
3，虚函数的调用一律通过vtable定位，因为存在vtable这一层中继，就可以保证在运行态可以通过父类的指针执行指向子类对象的方法，因为类对象初始化的时候正确的重写了vtable的内容。
4，一个对象多一个包含virtual function的基类，就会在类变量的结构中多引入一个指向父类vtable structure的指针。在类对象初始化的时候，就需要更多的 vtable 初始化代码。:D

多重继承会导致 vtable 膨胀， Microsoft ATL 引入了一个 ATL_NO_VTABLE __declspec(novtable) C++ 扩展。关于这个，下回再说~~~

其实这个例子里面还有一个小问题, 不知道大家能不能看出来啊!

Nortel files for bankruptcy

百年老店终于也挺不住了， NT的问题是积重难返，如果没有本质性的改变，估计会就此消失了。即便能恢复回来，估计也就不剩什么了~~~ :(

VIM is so cool

VIM is really cool, you can increase/decrease a number in the text directly by with CTRL-A and CTRL-X!

Adding and subtracting 

       *CTRL-A*
CTRL-A   Add [count] to the number or alphabetic character at
   or after the cursor.  {not in Vi}


       *CTRL-X*
CTRL-X   Subtract [count] from the number or alphabetic
   character at or after the cursor.  {not in Vi}

The CTRL-A and CTRL-X commands work for (signed) decimal numbers, unsigned
octal and hexadecimal numbers and alphabetic characters.  This depends on the
'nrformats' option.
- When 'nrformats' includes "alpha", Vim will change the alphabetic character
under or after the cursor.  This is useful to make lists with an alphabetic
index.
- When 'nrformats' includes "hex", Vim assumes numbers starting with '0x' or
'0X' are hexadecimal.  The case of the rightmost letter in the number
determines the case of the resulting hexadecimal number.  If there is no
letter in the current number, Vim uses the previously detected case.
- When 'nrformats' includes "octal", Vim considers numbers starting with a '0'
to be octal.  Other numbers are decimal and may have a preceding minus sign.
If the cursor is on a number, the commands apply to that number; otherwise
Vim uses the number to the right of the cursor.

For numbers with leading zeros (including all octal and hexadecimal numbers),
Vim preserves the number of characters in the number when possible.  CTRL-A on
"0077" results in "0100", CTRL-X on "0x100" results in "0x0ff".  Note that
when 'nrformats' includes "octal", decimal numbers with leading zeros are
impossible because they are indistinguishable from octal numbers.

The CTRL-A command is very useful in a macro.  Example: Use the following
steps to make a numbered list.

1. Create the first list entry, make sure it starts with a number.
2. qa      - start recording into buffer 'a'
3. Y      - yank the entry
4. p      - put a copy of the entry below the first one
5. CTRL-A    - increment the number
6. q      - stop recording
7. @a - repeat the yank, put and increment  times

Mapping remote TCP port to local via HTTPS Proxy

上次说的Perl代码供大家参考

功能：把远端的应用服务端口通过HTTPS代理服务器映射成本地端口。这对于那些不支持HTTPS代理的TCP应用还有点儿小用处！ :D

#!/usr/bin/env perl
#
# Use remote HTTPS PROXY to tunnel the remote service to local!
# Author: Yi Zhao
#   Blog: linuxyz.blogspot.com
#
# Based on original "ssltunnel.pl" by Alex Hornby <alex@hornby.org.uk>
# $Id: ssltunnel.pl,v 1.17 2003/06/10 14:54:16 alex Exp $
#
# This program is free software; you can redistribute it and/or
# modify it under the same terms as Perl itself.
#

package httpsproxy;

use strict;
my $VERSION=1.0;

use IO::File;
use IO::Select;
use IO::Socket;
use Net::SSL;
use Getopt::Long;
use MIME::Base64;

my %options = (
    proxyport=>443,
    reproxyport=>8080,
    localaddr=>"127.0.0.1",
    localport=>8080,
);

sub usage
{
    print STDERR <<EOF;
usage: perl httpsproxy.pl [options] [remote-host:port] 
Tunnels a TCP/IP connection through an http proxy using SSL.

WARNING: Only use this if you have the proxy administrator\'s permission
WARNING: The authors of this package offer no warranty

options:
    --help                        This help
    --proxyhost 1.2.3.4           Mandatory proxy host port (default 443)
    --proxyport 443               Optional proxy host port (default 443)
    --proxyuser <username>        Optional proxy user name
    --proxypasswd pass            Optional proxy pass word
    --useragent agent             Optional user agent name
    --reproxyhost 1.2.3.4         Optional intermediate http proxy host
    --reproxyport 123             Optional intermediate http proxy host port (default 8080)
    --reproxyuser <username>      Optional intermediate http proxy user name
    --reproxypasswd pass          Optional intermediate http proxy pass word
    --localaddr 1.2.3.4           Optional local addr to listen on (default 127.0.0.1)
    --localport 123               Optional local port to listen on (default 8080)

e.g. To run a local http proxy to the remote HTTPS proxy

./httpsproxy.pl --proxyhost 10.1.1.80 \
  --proxyport 443 --localport 80  www.abc.com:80

EOF
    exit(1);
}

# Parse command line arguments
sub parseArgs
{
    GetOptions(\%options, qw/dumpfile=s 
                   proxyhost=s proxyport=i proxyuser=s proxypasswd=s useragent=s
                   reproxyhost=s reproxyport=i reproxyuser=s reproxypasswd=s reuseragent=s
                   localaddr=s localport=i 
                   debug! help!
                   log-file=s pidfile=s/);

    if ( $options{help} ) {
        usage();
    }

    for (@ARGV) {
        if (m/^([^:]*):(.*)$/) {
            $options{desthost} = $1;
            $options{destport} = int($2) || getservbyname($2, 'tcp') || die "unknown port";
            print "Mapping $options{desthost}:$options{destport} as $options{localaddr}:$options{localport}\n";
        }
    }
    
    if ( !defined($options{proxyhost}) ) {
        print STDERR "error: You must give a proxyhost\n";
        usage();
    }
}

# This is really inefficient, but we only use it for reading the proxy response
# so that does not really matter.
sub xgetline($)
{
    my $proxy = shift;
    my $val="";
    my $buf;
    do {
        $proxy->read($buf, 1);
        $val .= $buf;
    } until ($buf eq "\n");

    $val;
}

sub httpProxy
{
    my ( $proxy, $desthost, $destport, $proxyuser, $proxypasswd ) = @_;
    # Force flushing of socket buffers

    # The actual connect
    $proxy->print("CONNECT " .  $desthost . ":" .
                      $destport . " HTTP/1.0\r\n");
    if ( $options{debug} ) {
        print STDERR "CONNECT " .  $desthost . ":" .
                      $destport . " HTTP/1.0\n";
    }

    # Basic auth if needed
    if ( $proxyuser ) {
        my $auth = encode_base64(
            $proxyuser . ":" . $proxypasswd);
        $proxy->print("Proxy-authorization: Basic $auth\r\n");
        if ( $options{debug} ) {
            print STDERR "Proxy-authorization: Basic $auth"
        }
    }

    # User agent name if needed
    if ( $options{useragent} ) {
        $proxy->print("User-Agent: " . $options{useragent} . "\r\n");
        if ( $options{debug} ) {
            print STDERR "User-Agent: " . $options{useragent} . "\n";
        }
    }

    # end of headers
    $proxy->print("\r\n");

    my $status;

    # Wait for HTTP status code, bail out if you don't get back a 2xx code.
    #$_ = $proxy->getline();
    #$_ = $proxy->getchunk();
    $_ = xgetline($proxy);
    next if /^[\r]*$/;
    ($status) = (split())[1];
    die("Received a bad status code \"$status\" from proxy server\n$_") 
        if ( int($status/100) != 2 );

    while($_ = xgetline($proxy)) {
        chomp;   # Strip <LF>
        last if /^[\r]*$/;                # Empty line or a single <CR> left

        if ( $options{debug} ) {
            print STDERR "Got extra data [$_]\n";
        }
    }

    return $status;
}

sub connectProxy
{
    if ( $options{reproxyhost} ) {
        # proxy support
        $ENV{HTTPS_PROXY} = qq(http://$options{reproxyhost}:$options{reproxyport});

        # proxy_basic_auth
        $ENV{HTTPS_PROXY_USERNAME} = $options{reproxyuser} if $options{reproxyuser};
        $ENV{HTTPS_PROXY_PASSWORD} = $options{reproxypass} if $options{reproxypass};
    }

    # debugging (SSL diagnostics)
    $ENV{HTTPS_DEBUG} = $options{debug} ? 1 : 0;
   
    my $proxy = new Net::SSL (
        PeerAddr => $options{proxyhost},
        PeerPort => $options{proxyport},
        Proto => 'tcp',
    );

    die "Error connecting to proxy host $options{proxyhost} " . 
        "port $options{proxyport}: $!\n" unless $proxy;

    # Force flushing of socket buffers
    $proxy->autoflush(1);

    # only if the remote host are speficied
    httpProxy($proxy, $options{desthost}, $options{destport},
        $options{proxyuser}, $options{proxypasswd} )
    if ( $options{desthost} );

    return $proxy;
}

sub connectLocal
{
    my $listen = new IO::Socket::INET (
        Listen=> 5,
        LocalAddr => $options{localaddr},
        LocalPort => $options{localport},
        Proto => 'tcp',
        Reuse => 1,
    );

    die "can't listen on " . $options{localaddr} . ":"
        . $options{localport} unless $listen;

    print STDERR "Accepting network clients on " .
        $options{localaddr} . ":" .$options{localport} . "\n";

    my %client2proxy;
    my %proxy2client;

    my $s = IO::Select->new();
    $s->add($listen);

    my $dumpfh;
    if ( $options{dumpfile} ) {
        $dumpfh = new IO::File($options{dumpfile}, "w")
            or die "could not open dump file $_";
        $dumpfh->autoflush(1);
    }

    while ( 1 ) {
        my @res = IO::Select::select($s, undef, undef, 3600);
        if ( @res == 0 ) {
            print STDERR "got select error\n";
            last;
        }
        my ($read, $write, $error) = @res;

        # Check for disconnect
        for my $fh ( @$error ) {
            print STDERR "socket $fh is in error\n";
            $s->remove($fh);
            exit();
        }

        # Process handles ready to read;
        for my $fh ( @$read  ) {

            if ( $fh == $listen ) {
                my $client = $listen->accept();
                $client->autoflush(1);
                $s->add($client);
                my $proxy = connectProxy();
                $s->add($proxy);
                print STDERR "New connection from " . $client->peerhost() . "\n";
                $client2proxy{$client} = $proxy;
                $proxy2client{$proxy} = $client;
            } else {
                my $destfh;
                my $isclient = 0;
                if ( exists( $client2proxy{$fh} ) ) {
                    $destfh = $client2proxy{$fh};
                    $isclient = 1;
                } elsif ( exists( $proxy2client{$fh} ) ) {
                    $destfh = $proxy2client{$fh};
                }

                my $num = sysread($fh, $_, 4096);
                if ( $num) {
                    syswrite($destfh, $_, $num);
                    # Optional dump of traffic
                    if ( $dumpfh ) {
                        if ($isclient) {
                            $dumpfh->print("client[$_]\n");
                        } else {
                            $dumpfh->print("proxy[$_]\n");
                        }
                    }
                } else {
                    $s->remove($fh);
                    $s->remove($destfh);
                    if ( $isclient ) {
                        print STDERR "client disconnected\n";
                        delete($client2proxy{$fh});
                        delete($proxy2client{$destfh});
                    } else {
                        print STDERR "proxy disconnected\n";
                        delete($client2proxy{$destfh});
                        delete($proxy2client{$fh});
                    }
                    close($fh);
                    close($destfh);

                    if(%proxy2client == 0 ) {
                        print STDERR "last client disconnected\n";
                    }
                }
            }
        }
    }
}

parseArgs();
connectLocal();

0;

__END__

copy and paste in VIM

VIM have multiple clipboard which allows you copy and save different content at the same time. The register can be any of {a-zA-Z0-9.%#:-"}. It is really interesting thing which I never notice before.

"{a-zA-Z0-9.%#:-"}   Use register {a-zA-Z0-9.%#:-"} for next delete, yank
                  or put (use uppercase character to append with
                  delete and yank) ({.%#:} only work with put).

:reg[isters]         Display the contents of all numbered and named
                  registers.  {not in Vi}

:reg[isters] {arg}   Display the contents of the numbered and named
                  registers that are mentioned in {arg}.  For example:

["x]y{motion}        Yank {motion} text [into register x].  When no
                  characters are to be yanked (e.g., "y0" in column 1),
                  this is an error when 'cpoptions' includes the 'E'
                  flag.

["x]p                Put the text [from register x] after the cursor
                  [count] times.  {Vi: no count}

I forgot there is another register *, it is as same as the system clipboard. So you can yank some text into register *, it will be in the system clipboard also. :D