Redirect
The ICMP Redirect type is sent in a single case. Consider this, you have a network (192.168.0.0/24) with several clients and hosts on it, and two gateways. One gateway to a 10.0.0.0/24 network, and a default gateway to the rest of the Internet. Now consider if one of the hosts on the 192.168.0.0/24 network has no route set to 10.0.0.0/24, but it has the default gateway set. It sends a packet to the default gateway, which of course knows about the 10.0.0.0/24 network. The default gateway can deduce that it is faster to send the packet directly to the 10.0.0.0/24 gateway since the packet will enter and leave the gateway on the same interface. The default gateway will hence send out a single ICMP Redirect packet to the host, telling it about the real gateway, and then sending the packet on to the 10.0.0.0/24 gateway. The host will now know about the closest 10.0.0.0/24 gateway, and hopefully use it in the future.
The main header of the Redirect type is the Gateway Internet Address field. This field tells the host about the proper gateway, which should really be used. The packet also contains the IP header of the original packet, and the 64 first bits of data in the original packet, which is used to connect it to the proper process sending the data.
The Redirect type has 4 different codes as well, these are the following.
• Code 0 - Redirect for network - Only used for redirects for a whole network (e.g., the example above).
• Code 1 - Redirect for host - Only used for redirects of a specific host (e.g., a host route).
• Code 2 - Redirect for TOS and network - Only used for redirects of a specific Type of Service and to a whole network. Used as code 0, but also based on the TOS.
• Code 3 - Redirect for TOS and host - Only used for redirects of a specific Type of Service and to a specific host. Used as code 1, but also based on the TOS in other words.
TTL equals 0
The TTL equals 0 ICMP type is also known as Time Exceeded Message and has type 11 set to it, and has 2 ICMP codes available. If the TTL field reaches 0 during transit through a gateway or fragment reassembly on the destination host, the packet must be discarded. To notify the sending host of this problem, we can send a TTL equals 0 ICMP packet. The sender can then raise the TTL of outgoing packets to this destination if necessary.
The packet only contains the extra data portion of the packet. The data field contains the Internet header plus 64 bits of the data of the IP packet, so that the other end may match the packet to the proper process. As previously mentioned, the TTL equals 0 type can have two codes.
• Code 0 - TTL equals 0 during transit - This is sent to the sending host if the original packet TTL reached 0 when it was forwarded by a gateway.
• Code 1 - TTL equals 0 during reassembly - This is sent if the original packet was fragmented, and TTL reached 0 during reassembly of the fragments. This code should only be sent from the destination host.
Parameter problem
The parameter problem ICMP uses type 12 and it has 2 codes that it uses as well. Parameter problem messages are used to tell the sending host that the gateway or receiving host had problems understanding parts of the IP headers such as errors, or that some required options where missing.
The parameter problem type contains one special header, which is a pointer to the field that caused the error in the original packet, if the code is 0 that is. The following codes are available:
• Code 0 - IP header bad (catchall error) - This is a catchall error message as discussed just above. Together with the pointer, this code is used to point to which part of the IP header contained an error.
• Code 1 - Required options missing - If an IP option that is required is missing, this code is used to tell about it.
Timestamp request/reply
The timestamp type is obsolete these days, but we bring it up briefly here. Both the reply and the request has a single code (0). The request is type 13 while the reply is type 14. The timestamp packets contains 3 32-bit timestamps counting the milliseconds since midnight UT (Universal Time).
The first timestamp is the Originate timestamp, which contains the last time the sender touched the packet. The receive timestamp is the time that the echoing host first touched the packet and the transmit timestamp is the last timestamp set just previous to sending the packet.
Each timestamp message also contains the same identifiers and sequence numbers as the ICMP echo packets.
Information request/reply
The information request and reply types are obsolete since there are protocols on top of the IP protocol that can now take care of this when necessary (DHCP, etc). The information request generates a reply from any answering host on the network that we are attached to.
The host that wishes to receive information creates a packet with the source address set to the network we are attached to (for example, 192.168.0.0), and the destination network set to 0. The reply will contain information about our numbers (netmask and ip address).
The information request is run through ICMP type 15 while the reply is sent via type 16.
SCTP Characteristics
Stream Control Transmission Protocol (SCTP) is a relatively new protocol in the game, but since it is growing in usage and complements the TCP and UDP protocols, I have chosen to add this section about it. It has an even higher reliability than TCP, and at the same time a lower overhead from protocol headers.
SCTP has a couple of very interesting features that can be interesting. For those who wish to learn more about this, read the RFC 3286 - An Introduction to the Stream Control Transmission Protocol and RFC 2960 - Stream Control Transmission Protocol document. The first document is an introduction to SCTP and should be very interesting to people who are still in need of more information. The second document is the actual specification for the protocol, which might be less interesting unless you are developing for the protocol or are really interested.
The protocol was originally developed for Telephony over IP, or Voice over IP (VoIP), and has some very interesting attributes due to this. Industry grade VoIP requires very high reliability for one, and this means that a lot of resilience has to be built into the system to handle different kind of problems. The following is a list of the basic features of SCTP.
• Unicast with Multicast properties. This means it is a point-to-point protocol but with the ability to use several addresses at the same end host. It can in other words use different paths to reach the end host. TCP in comparison breaks if the transport path breaks, unless the IP protocol corrects it.
• Reliable transmission. It uses checksums and SACK to detect corrupted, damaged, discarded, duplicated and reordered data. It can then retransmit data as necessary. This is pretty much the same as TCP, but SCTP is more resilient when it comes to reordered data and allows for faster pickups.
• Message oriented. Each message can be framed and hence you can keep tabs on the structure and order of the datastream. TCP is byte oriented and all you get is a stream of bytes without any order between different data inside. You need an extra layer of abstraction in TCP in other words.
• Rate adaptive. It is developed to cooperate and co-exist with TCP for bandwidth. It scales up and down based on network load conditions just the same as TCP. It also has the same algorithms for slow starting when packets where lost. ECN is also supported.