The Internet protocol suite providesend-to-end data communication specifying how data should be packetized, addressed, transmitted,routed, and received. This functionality is organized into fourabstraction layers, which classify all related protocols according to each protocol's scope of networking.[1][2] An implementation of the layers for a particular application forms aprotocol stack. From lowest to highest, the layers are thelink layer, containing communication methods for data that remains within a single network segment (link); theinternet layer, providinginternetworking between independent networks; thetransport layer, handling host-to-host communication; and theapplication layer, providing process-to-process data exchange for applications.
Thetechnical standards underlying the Internet protocol suite and its constituent protocols are maintained by theInternet Engineering Task Force (IETF). The Internet protocol suite predates theOSI model, a more comprehensive reference framework for general networking systems.
Initially referred to as theDOD Internet Architecture Model, the Internet protocol suite has its roots in research and development sponsored by the Defense Advanced Research Projects Agency (DARPA) in the late 1960s.[3] After DARPA initiated the pioneeringARPANET in 1969,Steve Crocker established a "Networking Working Group" which developed a host-host protocol, theNetwork Control Program (NCP).[4] In the early 1970s, DARPA started work on several other data transmission technologies, including mobile packet radio, packet satellite service, local area networks, and other data networks in the public and private domains. In 1972,Bob Kahn joined the DARPAInformation Processing Technology Office, where he worked on both satellite packet networks and ground-based radio packet networks, and recognized the value of being able to communicate across both. In the spring of 1973,Vinton Cerf joined Kahn with the goal of designing the next protocol generation for the ARPANET to enableinternetworking.[5][6] They drew on the experience from the ARPANET research community, theInternational Network Working Group, which Cerf chaired, and researchers atXerox PARC.[7][8][9]
By the summer of 1973, Kahn and Cerf had worked out a fundamental reformulation, in which the differences between local network protocols were hidden by using a commoninternetwork protocol, and, instead of the network being responsible for reliability, as in the existing ARPANET protocols, this function was delegated to the hosts. Cerf creditsLouis Pouzin andHubert Zimmermann, designers of theCYCLADES network, with important influences on this design.[10][11] The new protocol was implemented as theTransmission Control Program in 1974 by Cerf,Yogen Dalal and Carl Sunshine.[12]
Initially, the Transmission Control Program, the precursor to the later protocol suite, provided only areliable byte stream service, notdatagrams.[13] Several versions were developed by communication via theInternet Experiment Note series.[14] As experience with the protocol grew, collaborators recommended division of functionality into layers of distinct protocols, providing direct access to datagram service. Advocates includedBob Metcalfe and Yogen Dalal at Xerox PARC;[15][16]Danny Cohen, who needed it for hispacket voice work; andJonathan Postel of the University of Southern California'sInformation Sciences Institute, who edited theRequest for Comments (RFCs), the technical and strategic document series that has both documented and catalyzed Internet development.[17] Postel stated, "We are screwing up in our design of Internet protocols by violating the principle of layering."[18] Encapsulation of different mechanisms was intended to create an environment where the upper layers could access only what was needed from the lower layers. A monolithic design would be inflexible and lead to scalability issues. Inversion 4, written in 1978, Postel split the Transmission Control Program into two distinct protocols, theInternet Protocol as a connectionless layer and theTransmission Control Protocol as a reliableconnection-oriented service.[19][20][21][nb 1]
The design of the network included the recognition that it should provide only the functions of efficiently transmitting and routing traffic between end nodes and that all other intelligence should be located at the edge of the network, in the end nodes. Thisend-to-end principle was pioneered by Louis Pouzin in the CYCLADES network,[22] based on the ideas ofDonald Davies.[23][24] Using this design, it became possible to connect other networks to the ARPANET that used the same principle, irrespective of other local characteristics, thereby solving Kahn's initial internetworking problem. A popular expression is that TCP/IP, the eventual product of Cerf and Kahn's work, can run over "two tin cans and a string."[25] Years later, as ajoke in 1999, theIP over Avian Carriers formal protocol specification was created[26] and successfully tested two years later. Ten years later still, it was adapted for IPv6.[27]
DARPA contracted withBBN Technologies,Stanford University, andUniversity College London to develop operational versions of the protocol on several hardware platforms.[28] During development of the protocol the version number of the packet routing layer progressed from version 1 to version 4, the latter of which was installed in the ARPANET in 1983. It became known asInternet Protocol version 4 (IPv4), and it, along with its successor,Internet Protocol version 6 (IPv6), are thenetwork layer protocols used on the Internet.
In 1975, a two-network IP communications test was performed between Stanford and University College London. In November 1977, a three-network IP test was conducted between sites in the US, the UK, andNorway. Several other IP prototypes were developed at multiple research centers between 1978 and 1983.[14]
A computer called arouter is provided with an interface to each network. It forwardsnetwork packets back and forth between them.[29] Originally a router was calledgateway, but the term was changed to avoid confusion with other types ofgateways.[30]
In March 1982, the US Department of Defense declared TCP/IP as the standard for all military computer networking.[31][32][33] In the same year,NORSAR/NDRE andPeter Kirstein's research group atUniversity College London adopted the protocol.[34] The migration of the ARPANET fromNCP to TCP/IP was officially completed onflag day January 1, 1983, when the new protocols were permanently activated.[31][35]
In 1985, the Internet Advisory Board (laterInternet Architecture Board) held a three-day TCP/IP workshop for the computer industry, attended by 250 vendor representatives, promoting the protocol and leading to its increasing commercial use. In 1985, the firstInterop conference focused on network interoperability by broader adoption of TCP/IP. The conference was founded by Dan Lynch, an early Internet activist. From the beginning, large corporations, such as IBM and DEC, attended the meeting.[36][37]
IBM, AT&T and DEC were the first major corporations to adopt TCP/IP, this despite having competingproprietary protocols. In IBM, from 1984,Barry Appelman's group did TCP/IP development. They navigated the corporate politics to get a stream of TCP/IP products for various IBM systems, includingMVS,VM, andOS/2. At the same time, several smaller companies, such asFTP Software and theWollongong Group, began offering TCP/IP stacks forDOS andMicrosoft Windows.[38] The firstVM/CMS TCP/IP stack came from the University of Wisconsin.[39]
Some programmers are notable for early TCP/IP stack implementations. Jay Elinsky and Oleg Vishnepolsky of IBM Research wrote software for VM/CMS and OS/2, respectively.[40] In 1984, Donald Gillies atMIT wrote antcp multi-connection TCP which runs atop the IP/PacketDriver layer maintained by John Romkey at MIT in 1983–84. Romkey leveraged this TCP in 1986 when FTP Software was founded.[41][42] Starting in 1985, Phil Karn created a multi-connection TCP application for ham radio systems (KA9Q TCP).[43]
The spread of TCP/IP was fueled further in June 1989, when theUniversity of California, Berkeley agreed to place the TCP/IP code developed forBSD UNIX into the public domain. Various corporate vendors, including IBM, included this code in commercial TCP/IP software releases. For Windows 3.1, the dominant PC operating system among consumers in the first half of the 1990s, Peter Tattam's release of theTrumpet Winsock TCP/IP stack was key to bringing the Internet to home users. Trumpet Winsock allowed TCP/IP operations over a serial connection (SLIP orPPP). The typical home PC of the time had an external Hayes-compatible modem connected via an RS-232 port with an8250 or16550 UART which required this type of stack. Later, Microsoft would release their own TCP/IP add-on stack forWindows for Workgroups 3.11 and a native stack in Windows 95. These events helped cement TCP/IP's dominance over other protocols on Microsoft-based networks, which included IBM'sSystems Network Architecture (SNA), and on other platforms such asDigital Equipment Corporation'sDECnet,Open Systems Interconnection (OSI), andXerox Network Systems (XNS).
Nonetheless, for a period in the late 1980s and early 1990s, engineers, organizations and nations werepolarized over the issue of which standard, the OSI model or the Internet protocol suite, would result in the best and most robust computer networks.[44][45][46]
The characteristic architecture of the Internet protocol suite is its broad division into operating scopes for the protocols that constitute its core functionality. The defining specifications of the suite are RFC 1122 and 1123, which broadly outlines fourabstraction layers (as well as related protocols); the link layer, IP layer, transport layer, and application layer, along with support protocols.[1][2] These have stood the test of time, as the IETF has never modified this structure. As such a model of networking, the Internet protocol suite predates the OSI model, a more comprehensive reference framework for general networking systems.[46]
Conceptual data flow in a simple network topology of two hosts (A andB) connected by a link between their respective routers. The application on each host executes read and write operations as if the processes were directly connected to each other by some kind of data pipe. After establishment of this pipe, most details of the communication are hidden from each process, as the underlying principles of communication are implemented in the lower protocol layers. In analogy, at the transport layer the communication appears as host-to-host, without knowledge of the application data structures and the connecting routers, while at the internetworking layer, individual network boundaries are traversed at each router.Encapsulation of application data descending through the layers described in RFC 1122
Theend-to-end principle has evolved over time. Its original expression put the maintenance of state and overall intelligence at the edges, and assumed the Internet that connected the edges retained no state and concentrated on speed and simplicity. Real-world needs for firewalls, network address translators, web content caches and the like have forced changes in this principle.[49]
Therobustness principle states: "In general, an implementation must be conservative in its sending behavior, and liberal in its receiving behavior. That is, it must be careful to send well-formed datagrams, but must accept any datagram that it can interpret (e.g., not object to technical errors where the meaning is still clear)."[50]: 23 "The second part of the principle is almost as important: software on other hosts may contain deficiencies that make it unwise to exploit legal but obscure protocol features."[1]: 13
Encapsulation is used to provide abstraction of protocols and services. Encapsulation is usually aligned with the division of the protocol suite into layers of general functionality. In general, an application (the highest level of the model) uses a set of protocols to send its data down the layers. The data is further encapsulated at each level.
An early pair of architectural documents,RFC1122 and1123, titledRequirements for Internet Hosts, emphasizes architectural principles over layering.[51] RFC 1122/23 are structured in sections referring to layers, but the documents refer to many other architectural principles, and do not emphasize layering. They loosely defines a four-layer model, with the layers having names, not numbers, as follows:[1][2]
Theapplication layer is the scope within which applications, orprocesses, create user data and communicate this data to other applications on another or the same host. The applications make use of the services provided by the underlying lower layers, especially the transport layer which providesreliable or unreliablepipes to other processes. The communications partners are characterized by the application architecture, such as theclient–server model andpeer-to-peer networking. This is the layer in which all application protocols, such as SMTP, FTP, SSH, and HTTP, operate. Processes are addressed via ports which essentially representservices.
Thetransport layer performs host-to-host communications on either the local network or remote networks separated by routers.[52] It provides a channel for the communication needs of applications. TheUser Datagram Protocol (UDP) is the most basic[citation needed] transport layer protocol, providing an unreliableconnectionless datagram service. TheTransmission Control Protocol (TCP) provides flow-control, connection establishment, and reliable transmission of data.
Theinternet layer exchanges datagrams across network boundaries. It provides a uniform networking interface that hides the actual topology (layout) of the underlying network connections. It is therefore also the layer that establishes internetworking. Indeed, it defines and establishes the Internet. This layer defines the addressing and routing structures used for the TCP/IP protocol suite. The primary protocol in this scope is the Internet Protocol, which definesIP addresses.[53][failed verification][54] Its function in routing is to transport datagrams to the next host, functioning as an IP router, that has the connectivity to a network closer to the final data destination.[54][failed verification]
Thelink layer defines the networking methods within the scope of the local network link on which hosts communicate without intervening routers. This layer includes the protocols used to describe the local network topology and the interfaces needed to effect the transmission of internet layer datagrams to next-neighbor hosts.[55]
The protocols of the link layer operate within the scope of the local network connection to which a host is attached. This regime is called thelink in TCP/IP parlance and is the lowest component layer of the suite. The link includes all hosts accessible without traversing a router. The size of the link is therefore determined by the networking hardware design. In principle, TCP/IP is designed to be hardware independent and may be implemented on top of virtually any link-layer technology. This includes not only hardware implementations but also virtual link layers such asvirtual private networks andnetworking tunnels.
The link layer is used to move packets between the internet layer interfaces of two different hosts on the same link. The processes of transmitting and receiving packets on the link can be controlled in thedevice driver for thenetwork card, as well as infirmware or by specializedchipsets. These perform functions, such as framing, to prepare the internet layer packets for transmission, and finally transmit the frames to thephysical layer and over atransmission medium. The TCP/IP model includes specifications for translating the network addressing methods used in the Internet Protocol to link-layer addresses, such asmedia access control (MAC) addresses. All other aspects below that level, however, are implicitly assumed to exist and are not explicitly defined in the TCP/IP model.
The link layer in the TCP/IP model has corresponding functions in Layer 2 of the OSI model.
Internetworking requires sending data from the source network to the destination network. This process is calledrouting and is supported by host addressing and identification using the hierarchicalIP addressing system. The internet layer provides an unreliable datagram transmission facility between hosts located on potentially different IP networks by forwarding datagrams to an appropriate next-hop router for further relaying to its destination. The internet layer has the responsibility of sending packets across potentially multiple networks. With this functionality, the internet layer makes possible internetworking, the interworking of different IP networks, and it essentially establishes the Internet.
The Internet Protocol is the principal component of the internet layer, and it defines two addressing systems to identify network hosts and to locate them on the network. The original address system of theARPANET and its successor, the Internet, isInternet Protocol version 4 (IPv4). It uses a 32-bitIP address and is therefore capable of identifying approximately four billion hosts. This limitation was eliminated in 1998 by the standardization ofInternet Protocol version 6 (IPv6) which uses 128-bit addresses. IPv6 production implementations emerged in approximately 2006.
The transport layer establishes data channels that applications use for task-specific data exchange. The layer establishes host-to-host connectivity in the form of end-to-end message transfer services that are independent of the underlying network and independent of the structure of user data and the logistics of exchanging information. Connectivity at the transport layer can be categorized as eitherconnection-oriented, implemented in TCP, orconnectionless, implemented in UDP. The protocols in this layer may provideerror control,segmentation,flow control,congestion control, and application addressing (port numbers).
For the purpose of providing process-specific transmission channels for applications, the layer establishes the concept of thenetwork port. This is a numbered logical construct allocated specifically for each of the communication channels an application needs. For many types of services, theseport numbers have been standardized so that client computers may address specific services of a server computer without the involvement ofservice discovery ordirectory services.
Because IP provides only abest-effort delivery, some transport-layer protocols offer reliability.
TCP is a connection-oriented protocol that addresses numerous reliability issues in providing areliable byte stream:
data arrives in-order
data has minimal error (i.e., correctness)
duplicate data is discarded
lost or discarded packets are resent
includes traffic congestion control
The newerStream Control Transmission Protocol (SCTP) is also a reliable, connection-oriented transport mechanism. It is message-stream-oriented, not byte-stream-oriented like TCP, and provides multiple streams multiplexed over a single connection. It also providesmultihoming support, in which a connection end can be represented by multiple IP addresses (representing multiple physical interfaces), such that if one fails, the connection is not interrupted. It was developed initially for telephony applications (to transportSS7 over IP).
Reliability can also be achieved by running IP over a reliable data-link protocol such as theHigh-Level Data Link Control (HDLC).
TheUser Datagram Protocol (UDP) is a connectionlessdatagram protocol. Like IP, it is a best-effort, unreliable protocol. Reliability is addressed througherror detection using a checksum algorithm. UDP is typically used for applications such as streaming media (audio, video,Voice over IP, etc.) where on-time arrival is more important than reliability, or for simple query/response applications likeDNS lookups, where the overhead of setting up a reliable connection is disproportionately large.Real-time Transport Protocol (RTP) is a datagram protocol that is used over UDP and is designed for real-time data such asstreaming media.
The applications at any given network address are distinguished by their TCP or UDP port. By convention, certain well-known ports are associated with specific applications.
The TCP/IP model's transport or host-to-host layer corresponds roughly to the fourth layer in the OSI model, also called the transport layer.
QUIC is rapidly emerging as an alternative transport protocol. Whilst it is technically carried via UDP packets it seeks to offer enhanced transport connectivity relative to TCP.HTTP/3 works exclusively via QUIC.
The application layer includes the protocols used by most applications for providing user services or exchanging application data over the network connections established by the lower-level protocols. This may include some basic network support services such asrouting protocols and host configuration. Examples of application layer protocols include theHypertext Transfer Protocol (HTTP), theFile Transfer Protocol (FTP), theSimple Mail Transfer Protocol (SMTP), and theDynamic Host Configuration Protocol (DHCP).[56] Data coded according to application layer protocols areencapsulated into transport layer protocol units (such as TCP streams or UDP datagrams), which in turn uselower layer protocols to effect actual data transfer.
The TCP/IP model does not consider the specifics of formatting and presenting data and does not define additional layers between the application and transport layers as in the OSI model (presentation and session layers). According to the TCP/IP model, such functions are the realm oflibraries andapplication programming interfaces. The application layer in the TCP/IP model is often compared to a combination of the fifth (session), sixth (presentation), and seventh (application) layers of the OSI model.
Application layer protocols are often associated with particularclient–server applications, and common services havewell-known port numbers reserved by theInternet Assigned Numbers Authority (IANA). For example, theHyperText Transfer Protocol uses server port 80 andTelnet uses server port 23.Clients connecting to a service usually useephemeral ports, i.e., port numbers assigned only for the duration of the transaction at random or from a specific range configured in the application.
At the application layer, the TCP/IP model distinguishes betweenuser protocols andsupport protocols.[1]: §1.1.3 Support protocols provide services to a system of network infrastructure. User protocols are used for actual user applications. For example, FTP is a user protocol and DNS is a support protocol.
Although the applications are usually aware of key qualities of the transport layer connection such as the endpoint IP addresses and port numbers, application layer protocols generally treat the transport layer (and lower) protocols asblack boxes which provide a stable network connection across which to communicate. The transport layer and lower-level layers are unconcerned with the specifics of application layer protocols. Routers andswitches do not typically examine the encapsulated traffic, rather they just provide a conduit for it. However, somefirewall andbandwidth throttling applications usedeep packet inspection to interpret application data. An example is theResource Reservation Protocol (RSVP).[57] It is also sometimes necessary forApplications affected by NAT to consider the application payload.
Layering evolution and representations in the literature
The Internet protocol suite evolved through research and development funded over a period of time. In this process, the specifics of protocol components and their layering changed. In addition, parallel research and commercial interests from industry associations competed with design features. In particular, efforts in theInternational Organization for Standardization led to a similar goal, but with a wider scope of networking in general. Efforts to consolidate the two principal schools of layering, which were superficially similar, but diverged sharply in detail, led independent textbook authors to formulate abridging teaching tools.
The following table shows various such networking models. The number of layers varies between three and seven.
Some of the networking models are from textbooks, which are secondary sources that may conflict with the intent of RFC 1122 and otherIETF primary sources.[66]
The three top layers in the OSI model, i.e. the application layer, the presentation layer and the session layer, are not distinguished separately in the TCP/IP model which only has an application layer above the transport layer. While some pure OSI protocol applications, such asX.400, also combined them, there is no requirement that a TCP/IP protocol stack must impose monolithic architecture above the transport layer. For example, the NFS application protocol runs over theExternal Data Representation (XDR) presentation protocol, which, in turn, runs over a protocol calledRemote Procedure Call (RPC). RPC provides reliable record transmission, so it can safely use the best-effort UDP transport.
Different authors have interpreted the TCP/IP model differently, and disagree whether the link layer, or any aspect of the TCP/IP model, covers OSI layer 1 (physical layer) issues, or whether TCP/IP assumes a hardware layer exists below the link layer. Several authors have attempted to incorporate the OSI model's layers 1 and 2 into the TCP/IP model since these are commonly referred to in modern standards (for example, byIEEE andITU). This often results in a model with five layers, where the link layer or network access layer is split into the OSI model's layers 1 and 2.[67]
The IETF protocol development effort is not concerned with strict layering. Some of its protocols may not fit cleanly into the OSI model, although RFCs sometimes refer to it and often use the old OSI layer numbers. The IETF has repeatedly stated[47][failed verification] that Internet Protocol and architecture development is not intended to be OSI-compliant. RFC 3439, referring to the internet architecture, contains a section entitled: "Layering Considered Harmful".[66]
For example, the session and presentation layers of the OSI suite are considered to be included in the application layer of the TCP/IP suite. The functionality of the session layer can be found in protocols likeHTTP andSMTP and is more evident in protocols likeTelnet and theSession Initiation Protocol (SIP). Session-layer functionality is also realized with the port numbering of the TCP and UDP protocols, which are included in the transport layer of the TCP/IP suite. Functions of the presentation layer are realized in the TCP/IP applications with theMIME standard in data exchange.
Another difference is in the treatment ofrouting protocols. The OSI routing protocolIS-IS belongs to the network layer, and does not depend onCLNS for delivering packets from one router to another, but defines its own layer-3 encapsulation. In contrast,OSPF,RIP,BGP and other routing protocols defined by the IETF are transported over IP, and, for the purpose of sending and receiving routing protocol packets, routers act as hosts. As a consequence, routing protocols are included in the application layer.[29] Some authors, such as Tanenbaum inComputer Networks, describe routing protocols in the same layer as IP, reasoning that routing protocols inform decisions made by the forwarding process of routers.
IETF protocols can be encapsulated recursively, as demonstrated by tunnelling protocols such asGeneric Routing Encapsulation (GRE). GRE uses the same mechanism that OSI uses for tunnelling at the network layer.
^Cerf, Vinton G. & Cain, Edward (October 1983). "The DoD Internet Architecture Model".Computer Networks.7 (5). North-Holland:307–318.doi:10.1016/0376-5075(83)90042-9.
^Cerf, V.; Kahn, R. (1974)."A Protocol for Packet Network Intercommunication"(PDF).IEEE Transactions on Communications.22 (5):637–648.doi:10.1109/TCOM.1974.1092259.ISSN1558-0857.Archived(PDF) from the original on October 10, 2022. RetrievedOctober 18, 2015.The authors wish to thank a number of colleagues for helpful comments during early discussions of international network protocols, especially R. Metcalfe, R. Scantlebury, D. Walden, and H. Zimmerman; D. Davies and L. Pouzin who constructively commented on the fragmentation and accounting issues; and S. Crocker who commented on the creation and destruction of associations.
^"The internet's fifth man".Economist. December 13, 2013.Archived from the original on April 19, 2020. RetrievedSeptember 11, 2017.In the early 1970s Mr Pouzin created an innovative data network that linked locations in France, Italy and Britain. Its simplicity and efficiency pointed the way to a network that could connect not just dozens of machines, but millions of them. It captured the imagination of Dr Cerf and Dr Kahn, who included aspects of its design in the protocols that now power the internet.
^Pelkey, James. "8.3 CYCLADES Network and Louis Pouzin 1971-1972".Entrepreneurial Capitalism and Innovation: A History of Computer Communications 1968-1988. Archived fromthe original on June 17, 2021. RetrievedNovember 21, 2021.The inspiration for datagrams had two sources. One was Donald Davies' studies. He had done some simulation of datagram networks, although he had not built any, and it looked technically viable. The second inspiration was I like things simple. I didn't see any real technical motivation to overlay two levels of end-to-end protocols. I thought one was enough.
^Davies, Donald; Bartlett, Keith; Scantlebury, Roger; Wilkinson, Peter (October 1967).A Digital Communication Network for Computers Giving Rapid Response at remote Terminals(PDF). ACM Symposium on Operating Systems Principles.Archived(PDF) from the original on October 10, 2022. RetrievedSeptember 15, 2020.all users of the network will provide themselves with some kind of error control
^by Vinton Cerf, as told to Bernard Aboba (1993)."How the Internet Came to Be". Archived fromthe original on September 26, 2017. RetrievedSeptember 25, 2017.We began doing concurrent implementations at Stanford, BBN, and University College London. So effort at developing the Internet protocols was international from the beginning.
^Hauben, Ronda (2004)."The Internet: On its International Origins and Collaborative Vision".Amateur Computerist.12 (2). RetrievedMay 29, 2009.Mar '82 – Norway leaves the ARPANET and become an Internet connection via TCP/IP over SATNET. Nov '82 – UCL leaves the ARPANET and becomes an Internet connection.
^Tanenbaum, Andrew S. (January 1, 2003).Computer Networks. Prentice Hall PTR. p. 42.ISBN0-13-066102-3. RetrievedSeptember 12, 2016 – via Internet Archive.networks.
Douglas E. Comer (2001).Internetworking with TCP/IP – Principles, Protocols and Architecture. CET [i. e.] Computer Equipment and Trade.ISBN86-7991-142-9.
Joseph G. Davies; Thomas F. Lee (2003).Microsoft Windows Server 2003 TCP/IP Protocols and Services. Microsoft Press.ISBN0-7356-1291-9.
Forouzan, Behrouz A. (2003).TCP/IP Protocol Suite (2nd ed.). McGraw-Hill.ISBN978-0-07-246060-5.
W. Richard Stevens (1996).TCP/IP Illustrated, Volume 3: TCP for Transactions, HTTP, NNTP, and the UNIX Domain Protocols. Addison-Wesley.ISBN0-201-63495-3.
