习惯了TCP编程，认为UDP可以包办这些问题是错误的。一个UDP应用程序要承担可靠性方面的全部工作，包括报文的丢失、重复、时延、乱序以及连接失效等问题。

通常我们在可靠性好，传输时延小的局域网上开发测试，一些问题不容易暴露，但在大型互联网上却会出现错误。

UDP协议把递送的可靠性责任推到了上层即应用层，下面简单编写了几个类来专门处理两个问题：乱序和丢包。

四个类：DataPacket 类，PacketHeader类，PacketBody类 ，DataEntry类，位于同一个文件DataPacket .java中。

DataPacket 类相当于一个门面模式，提供给外部使用，通信数据也在这个类中处理。

package com.skysoft.pcks;

import java.io.*;
import java.net.*;
import java.util.*;

public class DataPacket {
InputStream is;
OutputStream os;
PacketHeader header;
PacketBody body;
ArrayList al;
public static final int DataSwapSize = 64532;

/**
* 在接收数据报使用
*/
public DataPacket() {
header = new PacketHeader();
body = new PacketBody();
al = new ArrayList();
}
/**
* 在发送数据报时使用,它调用报文分割操作.
* @param file String 硬盘文件
*/
public DataPacket(String file) {
this();
try {
is = new FileInputStream(file);
header.CalcHeaderInfo(is.available());
this.madeBody();
is.close();
//this.Gereratedata();
}
catch (FileNotFoundException ex) {
ex.printStackTrace();
}
catch (IOException ex1) {
ex1.printStackTrace();
}
}
/**
* 在发送数据报时使用,它调用报文分割操作.
* @param url URL url地址
*/
public DataPacket(URL url) {
this();
try {
//is = url.openStream();
URLConnection conn=url.openConnection();
is=conn.getInputStream();
int total=conn.getContentLength();
header.CalcHeaderInfo(total);
this.madeBody();
//System.out.println(total+":"+total);
is.close();
}
catch (IOException ex) {
ex.printStackTrace();
}
}
/**
* 为发送构造分组,使用PackageHeader处理了报头格式,并为分组编序号.
*/
private void madeBody() {
al.clear();
byte[] buffer;
DataEntry de;
for (int i = 0; i < header.fragmentcounter; i++) {
try {
ByteArrayOutputStream bos = new ByteArrayOutputStream();
//is.skip(i * body.BODY_BUFFER_SIZE);
header.ArrageSort(i);
de = new DataEntry(PacketBody.BODY_BUFFER_SIZE);
de.setSn(i);
de.setStreamsize(header.getStreamsize());
de.setFragmentcounter(header.getFragmentcounter());
if (header.isWTailFragment(i)) {
buffer = new byte[header.getMinfragment()];
is.read(buffer, 0, buffer.length);
header.setActByteSize(header.getMinfragment());
de.setActByteSize(header.getMinfragment());
}
else {
buffer = new byte[body.BODY_BUFFER_SIZE];
is.read(buffer, 0, buffer.length);
}
//System.out.println("length-------"+i+" "+body.getBody().length+" "+header.getMinfragment());
body.setBody(buffer);
//System.out.println("length:" + i + " " + header.toString());
bos.write(header.getByte(), 0, header.HEADER_BUFFER_SIZE);
bos.write(body.getBody(), 0, body.getBody().length);
de.setBytes(bos.toByteArray());
al.add(de);
}
catch (IOException ex) {
ex.printStackTrace();
}
}
}
/**
* 为发送构造分组,没有考虑报头格式,也没有为分组编序号.
*/
private void madeBody1() {
al.clear();
for (int i = 0; i < header.fragmentcounter; i++) {
try {
if (header.isWTailFragment(i))
is.read(body.getBody(), i * body.BODY_BUFFER_SIZE,
header.getMinfragment());
else
is.read(body.getBody(), i * body.BODY_BUFFER_SIZE,
body.BODY_BUFFER_SIZE);
ByteArrayOutputStream bos = new ByteArrayOutputStream();
bos.write(header.getByte(), 0, header.HEADER_BUFFER_SIZE);
bos.write(body.getBody(), header.HEADER_BUFFER_SIZE,
body.getBody().length);
al.add(bos);
}
catch (IOException ex) {
ex.printStackTrace();
}
}
}
/**
* 在接收到报文后,对此报文执行组装,并处理报文丢失和乱序情况.
* @param b1 byte[]
*/
public void Add(byte[] b1) {
byte[] buffer = (byte[]) b1.clone();
handlerText(buffer);
DataEntry de = new DataEntry(buffer, header.getActByteSize());
de.setSn(header.getSn());
de.setStreamsize(header.getStreamsize());
de.setFragmentcounter(header.getFragmentcounter());
al.add(de);
}
private void handlerText(byte[] buffer) {
ByteArrayOutputStream baos = new ByteArrayOutputStream();
baos.write(buffer, 0, header.HEADER_BUFFER_SIZE);
byte[] b=new byte[header.HEADER_BUFFER_SIZE];
System.arraycopy(buffer,0,b,0,b.length);
ByteArrayInputStream bais = new ByteArrayInputStream(baos.toByteArray());
InputStreamReader isr = new InputStreamReader(bais);
BufferedReader br = new BufferedReader(isr);
try {
header = new PacketHeader(br.readLine());
}
catch (Exception ex) {
ex.printStackTrace();
}
}

private String calFileSize(int size) {
return size / 1024 + "K";
}

public ArrayList getDataPackets() {
return al;
}
/**
* 是否接收完毕,通过序号是否等于最大段数来判断,这也许有问题,比如,正好是最后一个段丢失了,这样
* 这个包整个就丢失了.
* @return
*/
public boolean isFull() {
return this.header.getSn() == this.header.getFragmentcounter() - 1 ? true : false;
}
/**
* 判断是否只有一个段.
* @return
*/
public boolean isZero() {
return this.header.getSn() == 0 ? true : false;
}
/**
* 该函数执行报文组装,不考虑丢失的报文.
* @return
*/
private ByteArrayOutputStream fetchDataPackets() {
ByteArrayOutputStream bos = new ByteArrayOutputStream();
byte[] buffer = null;
DataEntry de;
for (int i = 0; i < al.size(); i++) {
try {
de = this.getSnData(i);
buffer = de.getByte();
if (header.getStreamsize() == de.getStreamsize()) {
bos.write(de.getByte(), header.HEADER_BUFFER_SIZE, de.getActByteSize());
System.out.println(de.toString() + " -- fetchDataPackets");
}
}
catch (Exception ex) {
ex.printStackTrace();
}
}
return bos;
}

/**
* 该函数执行报文组装,对于丢失的报文,写入空报文.
* @return ByteArrayOutputStream
*/
private ByteArrayOutputStream fetchDataPackets_sn() {
ByteArrayOutputStream bos = new ByteArrayOutputStream();
byte[] buffer;
DataEntry de;
for (int i = 0; i < header.getFragmentcounter(); i++) {
try {
de = this.getSnData(i);
if (de == null) {
de = seachDeData(i);
}
buffer = de.getByte();
//System.out.println(de.getSn() + ":" + i);
//handlerText(buffer);
//bos.write(buffer, header.HEADER_BUFFER_SIZE,
// buffer.length - header.HEADER_BUFFER_SIZE);
if (header.getStreamsize() == de.getStreamsize()) {
bos.write(de.getByte(), header.HEADER_BUFFER_SIZE,
de.getActByteSize());
//System.out.println(de.toString());
}
}
catch (Exception ex) {
ex.printStackTrace();
}
}
return bos;
}

/**
* 对缓冲的数据包进行排序处理,即按顺序提取同一帧的数据,如果没有找到该序号的帧,则返回空值.
* @param sn int 要找的帧序号.
* @return DataEntry
*/
private DataEntry getSnData(int sn) {
DataEntry de = null;
for (int i = 0; i < al.size(); i++) {
de = (DataEntry) al.get(i);
if (header.getStreamsize() == de.getStreamsize()) {
if (sn == de.getSn())
break;
else
de = null;
}
}
return de;
}

/**
* 按序号开始向前或者是向后寻找最近的帧片段,日后可以增加请求重发功能,通过开一个通信连接.
* @param sn int
* @return DataEntry
*/
private DataEntry seachDeData(int sn) {
DataEntry de = null;
int initvalue, minvalue = 10000;
DataEntry back, fore = null;
for (int i = 0; i < al.size(); i++) {
de = (DataEntry) al.get(i);
if (header.getStreamsize() == de.getStreamsize()) {
initvalue = Math.abs(de.getSn() - sn);
if (de.getFragmentcounter() != de.getSn() && initvalue < minvalue) {
minvalue = initvalue;
fore = de;
}
}
}
return fore;
}

/**
* 除去最后一帧外,随机抽取一帧.
* @return DataEntry
*/
private DataEntry seachDeData() {
DataEntry de = null;
for (int i = 0; i < al.size(); i++) {
de = (DataEntry) al.get(i);
System.out.println("sky ::::" + de.getFragmentcounter() + ":" + de.getSn() +
":" + i);
if (header.getStreamsize() == de.getStreamsize()) {
if (de.getFragmentcounter() != de.getSn()) {
break;
}
}
}
return de;
}
/**
* 生成组装完的结果数据.因为用图像来做测试,所以令其返回图像.
* @return Image
*/
public java.awt.Image Gereratedata() {
ByteArrayInputStream bis;
java.awt.image.BufferedImage bimage = null;
try {
byte[] b = fetchDataPackets_sn().toByteArray();
//fetchDataPackets_old1()
bis = new ByteArrayInputStream(b);
bimage = javax.imageio.ImageIO.read(bis);

}
catch (Exception ex1) {
ex1.printStackTrace();
}
return bimage;
}

public static void main(String args[]) {
DataPacket dp = new DataPacket("e:\\nature\\14.jpg");
}
}
/**
* 数据实体,充当临时处理场所.
* @author Administrator
*
*/
class DataEntry {
byte[] bytes;
int fragmentcounter, sn, actbytesize;
long streamsize;
int minfragment;

public DataEntry() {

}

public DataEntry(int size) {
this.actbytesize = size;
}

public DataEntry(byte[] b, int i) {
this.bytes = b;
this.actbytesize = i;
}

public byte[] getByte() {
return this.bytes;
}

public void setBytes(byte[] b) {
this.bytes = b;
}

public void setStreamsize(long size) {
this.streamsize = size;
}

public long getStreamsize() {
return this.streamsize;
}

public int getMinfragment() {
return minfragment;
}

public synchronized void setSn(int i) {
this.sn = i;
}

public synchronized int getSn() {
return sn;
}

public synchronized int getFragmentcounter() {
return fragmentcounter;
}

public synchronized void setFragmentcounter(int c) {
this.fragmentcounter = c;
}

public void setActByteSize(int size) {
actbytesize = size;
}

public int getActByteSize() {
return actbytesize;
}

public String toString() {
return this.streamsize + "::" + this.fragmentcounter + "::" + this.sn +
"::" + this.actbytesize + " recv DataEntry";
}
}
/**
* 报头,处理报头格式
* @author Administrator
*
*/
class PacketHeader implements Serializable{
public static final int HEADER_BUFFER_SIZE = 1024;
int fragmentcounter, sn;
int actbytesize = PacketBody.BODY_BUFFER_SIZE;
byte[] header; //= new byte[HEADER_BUFFER_SIZE];
long streamsize;
int minfragment;

public PacketHeader() {

}

public PacketHeader(long l) {
this.setStreamsize(l);

}

public PacketHeader(String s) {
String[] tm = s.split("::");
this.setActByteSize(Integer.parseInt(tm[3]));
this.setSn(Integer.parseInt(tm[2]));
this.setFragmentcounter(Integer.parseInt(tm[1]));
this.setStreamsize(Long.parseLong(tm[0]));
}

/**
* 根据文件的段的顺序生成数据头.
* @param sn 文件序列
*/
public void ArrageSort(int sn) {
this.setSn(sn);
this.setByte();
}

public void CalcHeaderInfo(long l) {
this.setStreamsize(l);
CalcHeaderInfo();
}
/**
* 计算流要被分成的片段数量,并得出最小片段余量.
*/
public void CalcHeaderInfo() {
fragmentcounter = Math.round( (float) streamsize /
PacketBody.BODY_BUFFER_SIZE);
float critical = (float) streamsize / PacketBody.BODY_BUFFER_SIZE;
if (critical - fragmentcounter < 0.5 && critical - fragmentcounter > 0)
fragmentcounter++;
minfragment = (int) (streamsize % PacketBody.BODY_BUFFER_SIZE);
}

public byte[] getHeader() {
Long it = new Long(this.streamsize);
return new byte[] {it.byteValue()};
}

public byte[] getByte() {
return header; //this.toString().getBytes();
}
/**
* 生成报头字节,首先取得数据包头 流尺寸::段片数::段顺序::段实际尺寸 的字节形式,
* 然后加入回车换行符号,对于1024字节中剩余的部分一律写入元素为0的字节数组.
*/
public void setByte() {
ByteArrayOutputStream bos = new ByteArrayOutputStream();
byte[] buffer = this.toByte();
try {
bos.write(buffer);
bos.write("rn".getBytes());
bos.write(new byte[PacketHeader.HEADER_BUFFER_SIZE - buffer.length], 0,
PacketHeader.HEADER_BUFFER_SIZE - buffer.length);
header = bos.toByteArray();
}
catch (IOException ex) {
ex.printStackTrace();
}
}

public void setStreamsize(long size) {
this.streamsize = size;
}

public long getStreamsize() {
return this.streamsize;
}

public int getMinfragment() {
return minfragment;
}

public synchronized void setSn(int i) {
this.sn = i;
}

public int getSn() {
return sn;
}

public int getFragmentcounter() {
return fragmentcounter;
}

public synchronized void setFragmentcounter(int c) {
this.fragmentcounter = c;
}

public void setActByteSize(int size) {
actbytesize = size;
setByte();
}

public int getActByteSize() {
return actbytesize;
}
/**
* 数据包头的格式为:流尺寸::段片数::段顺序::段实际尺寸
* 报头字节长度是可变化的,比如,可以加入流的具体信息如:流所属文件的名称,文件类型以及一些其他信息.
* @return String
*/
public String toString() {
return streamsize + "::" + this.fragmentcounter + "::" + this.getSn() +
"::" + this.getActByteSize();
}

public byte[] toByte() {
return this.toString().getBytes();
}
/**
* 是否为尾段
* @param i int
* @return boolean
*/
public boolean isWTailFragment(int i) {
return (i == fragmentcounter - 1) ? true : false;
}

}
/**
* 用户数据区
* @author Administrator
*
*/
class PacketBody implements Serializable{
public static final int BODY_BUFFER_SIZE = 63508; //65508
byte[] body;

public PacketBody() {
}

public void setBody(byte[] b) {
this.body = b;
}

public byte[] getBody() {
return body;
}
}

这个数据处理类，将在接下来使用。