This document covers the following topics:
With the J2SDK/JRE 1.4 release, IPv6 support has been added to Java Networking. This will make J2SE compliant with the following standards (RFCs):
Since the J2SDK does not support raw sockets, RFC 2292 (: Advanced Sockets API for IPv6) is not supported in this release.
Other interesting features of IPv6, such as tunneling, autoconfiguration of addresses, mobile IP, etc., are not supported at the Java API level, as they are handled automatically by the underlying OS or system support.
The following operating systems are supported in this release:
J2SETM 1.4 does not support IPv6 on Microsoft Windows. Microsoft has released an IPv6 Technology Previous for Windows 2000 but this is a limited seperate stack implementation. A complete dual IPv4/IPv6 stack implementation for Microsoft Windows is not expected to be released until 2002 or later.
This is also called anylocal or wildcard address. If a socket is bound to an IPv6 anylocal address on a dual-stack machine, it can accept both IPv6 and IPv4 traffic; if it is bound to an IPv4 (IPv4-mapped) anylocal address, it can only accept IPv4 traffic. We always try to bind to IPv6 anylocal address on a dual-stack machine unless a related system property is set to use IPv4 Stack.
When bound to ::, method ServerSocket.accept
will accept connections from both IPv6 or IPv4 hosts. The Java platform API
currently has no way to specify to accept connections only from IPv6 hosts.
Applications can enumerate the interfaces using NetworkInterface and bind a ServerSocketChannel to each IPv6 address, and then use a selector from the New I/O API to accept connections from these sockets.
|
Note: The option discussed below is introduced in Draft-ietf-ipngwg-rfc2553bis-03.txt. It will be supported in the Java 2 platform when it becomes a standard. |
However, there is a new socket option that changes the above behaviour. Draft-ietf-ipngwg-rfc2553bis-03.txt has introduced a new IP level socket option, IPV6_V6ONLY. This socket option restricts AF_INET6 sockets to IPv6 communications only. Normally, AF_INET6 sockets may be used for both IPv4 and IPv6 communications. Some applications may want to restrict their use of an AF_INET6 socket to IPv6 communications only. For these applications the IPV6_V6ONLY socket option is defined. When this option is turned on, the socket can be used to send and receive IPv6 packets only. By default this option is turned off.
Packets with the loopback address must never be sent on a link or forwarded by an IPv6 router. There are two separate loopback addresses for IPv4 and IPv6 and they are treated as such.
IPv4 and IPv6 addresses are separate address spaces except when it comes to "::".
This is used for hosts and routers to dynamically tunnel IPv6 packets over IPv4 routing infrastructure. It is meaningful for OS kernel and routers. Java provides a utility method to test it.
This is an IPv6 address that is used to represent an IPv4 address. It allows the native program to use the same address data structure and also the same socket when communicating with both IPv4 and IPv6 nodes. Thus, on a dual-stack node with IPv4-mapped address support, an IPv6 application can talk to both IPv4 and IPv6 peer. The OS will do the underlying plumbing required to send or receive an IPv4 datagram and to hand it to an IPv6 destination socket, and it will synthesize an IPv4-mapped IPv6 address when needed.
For Java, it is used for internal representation; it has no functional role. Java will never return an IPv4-mapped address. It understands IPv4-mapped address syntax, both in byte array and text representation. However, it will be converted into an IPv4 address.
On dual stack machines, system properties are provided for setting the preferred
protocol stack—IPv4 or IPv6—as well as the preferred
address family types—inet4 or inet6.
IPv6 stack is preferred by default, since on a dual-stack machine IPv6 socket can talk to both IPv4 and IPv6 peers.
This setting can be changed through the java.net.preferIPv4Stack=<true|false>
system property.
By default, we would prefer IPv4 addresses over IPv6 addresses, i.e., when querying the name service (e.g., DNS service), we would return Ipv4 addresses ahead of IPv6 addresses. There are two reasons for this choice:
This setting can be changed through the system property java.net.preferIPv6Addresses=<true|false>
For many years, if not forever, there will be a mix of IPv6 and IPv4 nodes on the Internet. Thus compatibility with the large installed base of IPv4 nodes is crucial for the success of the transition from IPv4 to IPv6. Dual stack, defined in RFC 1933, is one of the main mechanisms for guaranteeing a smooth transition. The other mechanism is IPv6 packet tunneling, which is relevant to the J2SDK only through the IPv4-compatible address. The former is the most relevant piece to the J2SDK. A dual stack includes implementations of both versions of the Internet Protocol, IPv4 and IPv6.
A general property of a dual-stack node is that an IPv6 socket can communicate both with an IPv4 and IPv6 peer at the transport layer (TCP or UDP) . At the native level, the IPv6 socket communicates with an IPv4 peer through an IPv4-mapped IPv6 address. However, unless a socket checks for the peers address type, it won't know whether it is talking to an IPv4 or an IPv6 peer. All the internal plumbing and conversion of address types is done by the dual-protocol stack.
Note: IPv4-mapped address has significance only at the implementation of a dual-protocol stack. It is used to fake (i.e., appear in the same format as) an Ipv6 address to be handed over to an IPv6 socket. At the conceptual level it has no role; its role is limited at the Java API level. Parsing of an IPv4-mapped address is supported, but an IPv4-mapped address is never returned.
| (Nodes) | V4 Only | V4/V6 | V6 Only |
|---|---|---|---|
| V4 Only | x | x | |
| V4/V6 | x | x | x |
| V6 Only | x | x |
Top row and left column represent various node types attempting to communicate. An x indicates that these nodes can communicate with each other.
Either host1 or host2 can be a native application.
If host2 wants to talk to host1, it will create a V6 socket. It then looks up the IP address for host1. Since host1 only has a v4 protocol stack, it will only have IPv4 records in the name lookup service. So host2 will try to reach host1 using an IPv4-mapped address. An IPv4 packet will be sent by host2, and host1 will think it is communicating with a v4 client.
If host2 is the server, it will first create a v6-type socket (by default), then it will wait for connections. Since host1 supports v4 only, it creates a v4-type socket. They resolves the name for host2. It only gets v4 address for host2, since it doesn't understand IPv6 address. So it connects to host2 using v4 address. A v4 packet will be sent on the wire. On host2, the dual stack will convert the v4 packet into a v6 packet with a v4-mapped address in it and hand it over to the v6 socket. The server application will handle it as if it is a connection from a v6 node.
InetAddressThis class represents an IP address. It provides address storage, name-address
translation methods, address conversion methods, as well as address testing
methods. In J2SE 1.4, this class is extended to support both IPv4 and IPv6 addresses.
Utility methods are added to check address types and scopes. The two types of
addresses, IPv4 and IPv6, can be distinguished by the Java type Inet4Address
and Inet6Address.
Two new subclasses of InetAddress are created: Inet4Address
and Inet6Address. V4- and V6-specific state and behaviors are implemented
in these two subclasses. Due to Java's object-oriented nature, an application
normally only needs to deal with InetAddress class—through
polymorphism it will get the correct behavior. Only when it needs to access
protocol-family-specific behaviors, such as in calling an IPv6-only method,
or when it cares to know the class types of the IP address, will it ever become
aware of Inet4Address and Inet6Address.
The new methods introduced are:
InetAddress:
isAnyLocalAddress
isLoopbackAddress
isLinkLocalAddress
isSiteLocalAddress
isMCGlobal
isMCNodeLocal
isMCLinkLocal
isMCSiteLocal
isMCOrgLocal
getCanonicalHostName
getByAddr
Inet6Address:
isIPv4CompatibleAddressInetAddress and Different Naming servicesPrior to 1.4, InetAddress utilized the system configured name
service to resolve host names. In 1.4, we have added a Java DNS provider through
JNDI for alternative name lookups. You can tell the J2SDK to use this provider
by setting up a few system properties. These system properties are documented
in the Java system properties section. In the future, we plan to provide a generic
service provider framework so that you can write your own name service providers.
All IPv4 addresses are represented in Java as Inet4Address objects.
They are serialized as InetAddress objects, and deserialized from
InetAddress to Inet4Address to keep backward compatibility.
IPv6 addresses are represented as Inet6Address and are serialized
as such.
Socket, ServerSocket, and DatagramSocketDue to the object-oriented nature of Java, the address types and storage structures are not exposed at the socket API level, so no new APIs are needed. The existing socket APIs handle both IPv4 and IPv6 traffic.
The selection of which stack to use depends upon the following:
All supported Ipv6 socket options have a IPv4 counterparts. Thus no new APIs were added to support IPv6 socket options. Instead, the old APIs are overloaded to support both V4 and V6 socket options.
MulticastSocket Again all the socket options APIs are overloaded to support IPv6 multicast socket options.
We have added two new APIs to set/get network interfaces in addition to the
existing MulticastSocket.setInterface/MulticastSocket.getInterface
that takes/returns an InetAddress instance. The two existing methods
are used to set or retrieve the network interface used by the current MulticastSocket
to send multicast packets (i.e., equivalent to IP_MULTICAST_IF
in native socket option). For IPv4, the interface was indicated by an IP address.
Thus we can use the equivalent InetAddress in Java. They will continue
to work with IPv6 multicast socket. However, in IPv6, according to RFC 2553,
the interface should be indicated using an interface index. To better support
the concept of a network interface, we introduced a new class, NetworkInterface.
It encapsulate the data representing the state of the network interface, including
name and IP addresses and some basic manipulation methods. Thus we have introduced
two new methods for setting the outgoing interface for multicast socket: setNetworkInterface
and getNetworkInterface. They take or return a NetworkInterface
object. These new methods can be used with both v4 and v6 multicast.
Methods have also been added for joining and leaving a multicast group on a network interface. This was previously unavailable in the Java API.
MulticastSocket:
NetworkInterface getNetworkInterface()
setNetworkInterface(NetworkInterface netIf)
joinGroup(SocketAddress mcastaddr,NetworkInterface netIf)
leaveGroup(SocketAddress mcastaddr,NetworkInterface netIf)
URL, URI parsersLiteral IP addresses can be used in URL/URIs. However, since colon (:)
is used in existing URL/URI specifications to separate host from port, using
literal IPv6 representation in URL/URIs without modification will fail in parsing.
Thus for specifying literal IPv6 addresses in URL/URIs, RFC 2732 was created.
The parsing of URL/URI has been updated to be compliant with RFC 2732.
Since SocketPermission utilizes URLs, its implementation has been
updated to be compliant with RFC 2732.
codebase, used in defining a permission, is a variant of URL.
As such, it should follow URL formats and conventions. RFC 2732 format is used
for URL and codebase; RFC 2373 format is used everywhere else.
java.net.preferIPv4Stack (default: false)If IPv6 is available on the operating system, the underlying native socket will be an IPv6 socket. This allows Java(tm) applications to connect too, and accept connections from, both IPv4 andIPv6 hosts.
If an application has a preference to only use IPv4 sockets, then this property can be set to true. The implication is that the application will not be able to communicate with IPv6 hosts.
java.net.preferIPv6Addresses (default: false)If IPv6 is available on the operating system, the default preference is to
prefer an IPv4-mapped address over an IPv6 address. This is for backward compatibility
reasons—for example, applications that depend on access to an IPv4-only
service, or applications that depend on the %d.%d.%d.%d representation of an
IP address.
This property can be set to try to change the preferences to use IPv6 addresses
over IPv4 addresses. This allows applications to be tested and deployed in environments
where the application is expected to connect to IPv6 services.
sun.net.spi.nameservice.provider.<n>=<default|dns,sun|...>Specifies the name service provider that you can use. By default, Java will
use the system-configured, name-lookup mechanism, such as file, nis, etc. You
can specify your own by setting this option. <n> takes the
value of a positive number and it indicates the precedence order: a small number
takes higher precendence over a bigger number. In 1.4 , we have provided one
DNS name service provider through JNDI, which is called dns,sun.
sun.net.spi.nameservice.nameservers=<server1_ipaddr,server2_ipaddr
...>You can specify a comma separated list of IP addresses that point to the DNS servers you want to use.
sun.net.spi.nameservice.domain=<domainname>This property specifies the default DNS domain name, e.g., eng.sun.com.