We offer a guide to the Internet "forest" which describes the players and the terms and technology they use. The guide is categorized by the various groups of technology that make up the Internet and all of the terms are arranged both alphabetically and in a narrative form. You can click on a specific term or abbreviation while reading the summary on our WEB pages and you will receive a definition of that term.

This glossary is organized by subject:

• Telephone Industry Terminology
• Data Networking Terminology
• Internet Specific Data Terminology
• Internet Service Provider Terminology
• Internet Related Organizations and Associated Terminology
• Internet Business Model Terminology

If you don't find what you are looking for, a very good glossary with a European flavor is available here.

The glossary is also available in alphabetical format.

Our glossary is now updateable and expandable by means of Perlscript which is contained in a custom designed Glossary Manager package by Andrey Vorobyov, Arcadia, Inc., St. Petersburg, Russia.

Telephone Industry Terminology

IXC

Inter Exchange Carrier is the post divestiture (1984) generic name for long distance phone companies in the United States'. AT&T is the largest, controlling more than 50% of the market. MCI and Sprint are the second and third largest IXC's. AT&T MCI, and Sprint are also international in scope. Several hundred more small IXC's exist. WilTel and LDDS were among the ten largest IXC’s before WilTel sold off its transmission business to LDDS. The resulting company became known as WorldCom and has embarked on an unprecedented acquisitions binge.

RBOC

The 1984 divestiture of AT&T left local telephone service under the control of seven Regional Bell Operating Companies (RBOCs), also sometimes referred to as RHCs (Regional Holding Companies). Each RBOC in turn is composed of several Local Exchange Carriers (LECs). The RBOCs and LECs operate under the same regulatory structures. The number of RBOCs is shrinking as the 1996 Telecommunications act has led to the merger of Pac Bell and Southwestern Bell, and Bell Atlantic and NYNEX.

LEC

The Local Exchange Carrier is the local telephone company for a given geographic area. In return for being given a monopoly over residential connections to the telephone network, the LEC, which is most likely one of the more than 20 former Bell pre-divestiture operating companies, has been subject to strict regulation of the services it offers and rates it may charge for those services. IXC's pay LECs a fee for termination of phone calls at the businesses and residences served by the LEC. With the advent of the 1996 Telecommunications Act LECs tend to come in two flavors: ILECs and CLECs. That is Incumbent Local Exchange Carrier and Competitive Local Exchange Carrier.

POTS

Plain Old Telephone Service refers to basic voice service available in residences throughout the United States for between 10 and 20 dollars a month.

MFJ

Modified Final Judgement is the name given Judge Green's decision outlining the rules of the 1984 divestiture of AT&T. Under the MFJ the RBOCs have been banned from manufacturing. Although allowed several years ago to provide information services, the RBOCs are still banned by the MFJ from delivery of inter-LATA telephone or data service. Again if the Telecom Deregulation Act of 1996 succedes in bringing competition to the local loop, the MFJ will dissappear. For the time being it is still in force.

LATA

The Local Access Transport Area was created by the 1984 divestiture. It defines the geographic area over which the Local Exchange Carrier may provide toll calls. The area is often smaller than that covered by a long distance area code. Even though ten or twenty LATAs are normally to be found within the territory of a Local Exchange Carrier, the LEC may not provide calls that cross LATA boundaries. Such inter-lata traffic is the exclusive domain of the Inter Exchange Carrier (IXC).

LDIP

Long Distance Internet Provider is the term coined to describe the alliance forced by the MFJ on RBOCs or LECs that wish to provide Internet service. Such companies must partner with an existing Internet Service Provider (ISP) which provides inter LATA backbone service to the local phone company points of presence (POPs).

Profit Cap Regulation

Before 1995 this was the predominant form of regulation applied to local phone companies in exchange for the monopoly of local phone service granted them by each state. Regulators looked at the total profit earned and, if it exceeded a 12 to 14% range, might force the phone company to give back excess profits with a rate reduction. In this world, the larger a company's gross revenues, the larger would be the base on which profits could be figured. Incentives to adopt new technologies and pare down the number of employees were not great.

Price Cap Regulation

By the end of 1995 a large number of the LECs had moved to Price Cap regulation where they promised not to raise rates for some specified period of time. They were now perfectly fee to earn larger profits and able to do so by adopting advanced digital technologies that allowed them to perform the same or improved services with far fewer employees. All five of Ameritech's LEC's were under this form of regulation by the end of 1995 and Ameritech profits were pushing into the 30% range.

Local loop

The local loop is often referred to as the last mile or last several miles from the IXC's lines to the customer's phones or modems. Operation of the local loop is the responsibility of the Local Exchange Carrier.

PSTN

The Public Switched Telephone Network (PSTN) refers to the combined infrastructure of the regulated IXC (AT&T) and the RBOCs and their respective Local Exchange Carriers. Universal telephone service embodied as the goal of the 1934 Communications Act is provided by access to the PSTN.

VAN

A Value Added Network is data network devoted to a specific application. Whereas the PSTN has been regulated since 1934, VANs are considered to be exempt from regulation.

Reciprocal Compensation for Internet Traffic

All RBOCs are now saying that they will not pay to terminate their end users calls to ISPs on a competitor's network. Under FTA96, (Federal Telecommunications Act of 1996) all LECs must pay reciprocal compensation to each other when one carrier's customer calls a customer served by another carrier and the call is "local." Under FCC rules, ISPs are treated as local end users. (Thanks to W Scott McCollough)

Comparably Efficient Interconnection

This is the filing and arrangement a RBOC must make with the FCC to get into enhanced/information services. They have to promise to make available - through tariff, and usually ONA (Open Network Architecture), a n unbundled group of network offerings through which competitors may have access to the features and functions necessary to do what the telco enhanced provider does. It is supposed to eliminate most of the incentive and ability to discriminate. The proces s is not working and the FCC is beginning to realize it. (Thanks to W. Scott McCollough)

Competitive Local Entry

The FCC has only two real sticks to keep the telcos from leveraging their monopoly power over the local network into competitive services. The first is competitive local entry, using FTA96 tools : A. unbundled network elements (similar in concept t o ONA); B. resale at a prescribed discount by competitors; C. removal of barriers to entry and construction of local networks by new entrants, with required interconnection with incumbents' networks and collocation at their premises. This is beginning to take hold, but will still take years to implement. (ThaNks to W. Scott McCollough)

Section 271 Authority

271 Authority is the FCC's second 'stick' for trying to enforce competition under the 1996 Telecom Act. It refers to FTA96 Section 271, which immediately allowed GTE into in-region inter LATA toll, but required the RBOCs to get FCC authority, by way of the states, with Department of Jus tice participation. The RBOCs must show they meet a prescribed laundry list of items, including the existence of actual competition in their local markets in each state. So far (October 1997) no RBOC has received the authority.

ILEC

ILEC is Incumbent Local Exchange Carrier. The RBOCs, also GTE, and Independent phone companies may be ILECs. ILECs are sometimes referred to as Dominant Carriers.

CLEC

CLEC is Competitive Local Exchange Carrier. Examples of CLECs are: MFS, Teleport, ACSI, AT&T (when they finally start), MCI Metro.

Data Networking Terminology

Internet

The Internet is a worldwide network of TCP/IP networks reaching into nearly 200 nations. Electronic mail can be gatewayed via the internet into other independent states and jurisdictions bringing the total to well over 200. The Internet now reaches every part of the globe.

Protocol

A protocol is the language that a network or network application "speaks." It is to networking what a programming language is to programming.

Protocol Stack

The protocol stack usually refers to the seven layer OSI protocol stack where TCP/IP occupies layers three and four and protocols, like x.25. Frame Relay and SMDS, level 2.

OSI

Open Systems Interconnection is set of data network architectural standards developed by ISO - the International Organization for Standardization. In the late 1970s and early 1980s it was thought that OSI standards would dominate data networking. However with the successful de-facto rise of the commercial Internet in the 1990s TCP/IP and associated IETF standards have become dominant.

ISO

International Organization for Standardization is the developer of the OSI standards.

GOSIP

Government OSI Profiles - a US government backed subset of the OSI standards.

TCP/IP

Transport Control Protocol/Internet Protocol (TCP/IP) has become, in a very short period, a world wide public domain standard for connecting computers by all vendors over wide area networks. It operates at level 3 and 4 of the 7 level protocol stack. Hence it can be transported by frame relay or SMDS which functions at level 2 of the stack.

SMTP

Simple Mail Transfer Protocol is the electronic mail or message transport protocol for the TCP/IP world and hence the Internet.

x.25

X.25 was the transport protocol for the earliest commercial packet data networks starting in the late 1970s. Speeds peaked at 56,000 bits per second

x.400

x.400 is the electronic mail protocol for OSI.

x.500

X.500 is the directory naming service protocol for OSI.

SNA

Systems Network Architecture is IBM's proprietary networking protocol used to enable its mainframes to communicate with each other. Before the development of multi-protocol routers, IBM could and did demand $80,000 to $300,000 for front end processors used in linking the mainframes. During the early nineties, many large companies abandoned these processors in favor of multi-protocol routers costing about $5,000 to 15,000 each and capable of encapsulating SNA traffic in TCP/IP packets.

DECnet

DECnet is Digital Equipment's proprietary networking protocol. It is facing the same problems that SNA is with customers desiring to use TCP/IP for networking all their equipment rather than having to run different dedicated networks for each different proprietary protocol.

EDI

Electronic Data Interchange is a set of standards that allows corporations to order from and send invoices to other corporations, all electronically by means of data networks.

Frame Relay

Frame relay is a level 2 fast packet switching service that takes up where the old x.25 networks left off (56 KBS) and goes to 45 megabits per second. The IXCs now have their own frame relay offerings which they will can to bring to businesses by means of virtual private networks (VPNs) that bypass the LECs entirely

Gateway

A gateway is an intersection between two networks running different protocols. A gateway router strips incoming packets of the protocol of the incoming network and encapsulates them in "envelopes" of the protocol of the outgoing network.

SMDS

Switched Multi-megabit Digital Service is a fast packet switching service. It can carry the TCP/IP protocol. The LECs however are at a disadvantage in that they will require a partnership with an IXC to carry SMDS data across LATA boundaries. . SMDS provides packet switched bandwidth, on demand, in increments up to 34 megabits.

ISDN

Integrated Services Digital Network technology has been available for more than a decade. Lack of equipment standardization and expensive modification for local computers have slowed down its deployment. In the last few years it has become widely available but, because most LECs insist on a per minute rather than flat charge, its growth in usage has been small. It can give a user up to 56 kilobits of data bandwidth on a phone line that is also used for voice or up to 128,000 bits per second if the user forgoes voice use of the ISDN line.

Asynchronous Transfer Mode or ATM

ATM came into widespread use during 1995. It is composed of 53 byte "cells" having 5 byte headers and 48 byte payloads. It is a high speed network protocol into which the phone companies have invested billions. Because of its short packet length, it is especially good for real time voice and video. Devotees of data networking scorn it because it wastes about 20% of all its bandwidth in the overhead of the 5 byte cell headers and in related overheads necessary for it to be a layer 2 transport service for TCP/IP.

Connection Oriented

The telephone network is connection oriented. This means that, for the duration of a telephone call, a small segment of the network is solely dedicated to the traffic of that one call. In other words no other calls can use that portion of the network.

Connectionless

Most computer data networks are connectionless. Data is encapsulated in "envelopes" called packets. The packets from a user's session may be sent by network routers along different routes to their destination as traffic conditions on the network change from moment to moment.

Packet switched

A Packet Switched Network is another term for connectionless data network. Data are inserted in packets which are the equivalent of software envelopes with addresses on them. These addresses can be read by routers, which by reference to internally contained routing tables can decide what network path to send the data on to ensure that it gets to its destination. The paths chosen by the routers can vary from moment to moment as each router gets updated information on the condition of other routers and circuits in the network.

Envelope

Envelope is another term for the data packet within which is held information desired by a network's end users.

Switched

Frame relay is a switched technology where packet headers need be only 2 bytes long. The first bytes of the switched network's protocol headers are composed of the permanent virtual circuit numbers necessary to direct data from one network node to another, plus a few control bits. Switching takes place at layer 2, routing at layer 3 of the seven layer stack. Switching is a much less CPU intensive activity than routing.

PVC

A Permanent Virtual Circuit is a connection oriented circuit that may be set up by software between any two nodes of a switched network.

Circuit

Circuit refers to a logical stream of data set up to flow through two or more network nodes. A single physical link between these nodes may have several virtual circuits flowing through it.

Routed

Routed is sometimes used to describe a network where data is routed at level 3 rather than switched at level 2. Parts of the Internet fabric maybe switched but some part of the fabric (network topology) between user and sender must always be routed.

Router

A Router is the device that serves as a "traffic cop" in a connectionless network such as the Internet. Routers are specialized computers that take incoming packets and compare their destination addresses to internal routing tables and, depending on network conditions, send the packets out to the appropriate receiving router. This process may be repeated many times until the packets reach their intended destination. The market for multi-protocol routers that include TCP/IP is one of the fastest growing within the telecommunications industry.

Cisco

Cisco is the largest maker of TCP/IP routers in the world. On major Internet backbones, large Cisco routers are almost invariably used - al though Bay Networks has begun to make some inroads.

VPN

A Virtual Private Network describes a network set up solely for the users of a single company. Such a network might have a gateway to the public Internet. With the deployment of IP sec (IP security) VPNs increasingly will ride on links of the public Internet.

Leased line

A leased line is the telephone circuit transmission channel reserved for the use of customer from point "a" to point "b" through phone company physical lines and switches. The line may be of different bandwidths of data carrying capacity. In data networking a bit pipe is a colloquial name for a leased line.

Bit Pipe

A bit pipe is the name given to a telephone circuit used for transmission of packets in a data network. A "dumb" bit pipe is a telephone circuit that provides only physical data layer transmission and no higher level applications.

Bandwidth

Bandwidth is the amount of data, measured usually in bits per second, that can be sent through a dedicated (leased) transmission circuit.

Bit

A bit is the primary unit of digital data. Written in binary language as a "1" or a "0".

Byte

A byte is composed of 8 bits.

ASCII character set

The ASCII character set refers to a uniform way of encoding bits into bytes, so that 128 differently coded bytes will each stand for a different letter, number, or punctuation mark in the Latin alphabet

56 Kilobit Leased Line

A 56 kilobit (56,000 bits per second) leased line is currently the smallest bandwidth transmission data circuit useful in Internet applications. It is also roughly the bandwidth needed for a voice phone call. Most Internet Service Providers now have T-1 lines (1.544 megabits per second).

T-1 circuit

A T-1 circuit or leased line equals 1,544,000 bits per second or 24 56 kbs leased lines.

E-1 circuit

An E-1 Circuit (2,000,000 bits per second) is the European equivalent (roughly speaking) of a T-1.

T-3

A T-3 Circuit (45,000,000 bits per second) was the backbone speed of all major national Internet service providers in the US. By the end of 1996 the largest backbones were running at OC-3. Some now (10/97) operate at OC-12. By late 1998 OC-12 should be the standard speed for major backbones.

OC-3

An OC-3 circuit (155,000,000 bits per second) is the backbone speed that major NSPs have upgraded their backbones to by the end of 1997.

OC-12

An OC-12 circuit (622,000,000 bits per second) is bandwidth that was experimented with in the Gigabit Testbeds of the early 1990s. At the beginning of 1998 it is also the bandwidth of Sprintlink and MCI's backbones. By the end of 98, it should equal the speed of every major NSP's backbone.

OC-48

An OC-48 circuit (2,400,000,000 bits or 2.4 gigabits per-second) is the typical speed for many aggregated telephone voice circuits on inter city fiber optic lines. Before the end of the decade most NSPs should be operating at OC-48 speeds. A few are expected to implement OC-48 before the end of 1998.

OC-192

An OC-192 circuit (9,600,000 bits or 9.6 gigabits per-second) is in late 1998 on the verge of becoming the standard inter-city trunks speed for next generation telcos (Qwest, Level 3, Williams, Frontier etc). With the advent of commercial Wave Division Multiplexing terrabit routers are now in betatest.

SONET

SONET or Synchronous Optical Network is a Bellcore-developed, CCITT, international standard for high speed communication over fiber-optic networks. SONET functions as a carrier for ATM fixed length packets (53 bytes). TCP/IP can ride on top of SONET and ATM. All major IXCs are completing the installation of SONET and ATM on their backbones at speeds ranging up to OC48 (2.4 gigabits per second). OC-3, etc refers to the measurements for SONET based circuit speed.

Wave Division Multiplexing (WDM)

Wave Division Multiplexing has come into commercial use in 1998. This a permits the creation of 16, or 32, or 96 and before long perhaps multiple hundreds of virtual fibers from a single glass thread. Lasers are used which transmit at a specific wavelength or lambda of light. These light waves can be multiplexed down a single fiber. One wave can be dedicated to IP traffic. Another to just ATM traffic and so on. IP can ride directly on WDM. SONET, in such a case, is not required. This has enormous cost implications. However because SONET is necessary to divide OC-48 or OC-192 lambdas into smaller bit sizes for resale and SONET is used for network redundancy, it won¹t go away overnight. So called optical networks depend on WDM.

Clear Channel Transmission

Clear Channel Transmission defines the amount of data occupied by a single user's network application. Al Gore's ideas about the need for an NREN not withstanding, clear channel gigabit TCP/IP transmission over a wide area network is not yet a proven technology let alone economically viable.

Aggregate Transmission

Aggregate Transmission refers to multiplexing or mixing together of the applications of thousands of users across a backbone. Such aggregate traffic can reach gigabit speeds with present technology and with acceptable dollar cost.

Twisted Pair

Twisted Pair refers to the standard two strands of copper that, with the appropriate insulation, have made up ordinary physical telephone lines for most of the 20th century. The data capacity of twisted pair is about 56 kilobits per second. However with special equipment and within a very few miles of phone company central offices speeds in excessive of several megabits per send data transmission have been achieved.

COAX

Coaxial cable is most often used in the home to attach to the back of a TV set to bring incoming cable TV signals to the set. The data capacity of coax can exceed 10 megabits per second.

Fiber

Fiber, or fiber optic, refers to cable containing often about two dozen threads of pure glass. Lasers attached to the end of such cable can send digital patterns of light pulses at hitherto unimaginable speeds. Compared to copper the carrying capacity of fiber for telecommunications signaling is almost unlimited.

Network topology

Network topology is a diagrammatic representation of the physical layout of the network. It includes a description of the hardware at the nodes and the structure adopted that will enable those links talk to each other.

Backbone

The backbone of a network is its means of linking its major nodes so that all its leaf nodes feed into backbone nodes with a high speed uninterrupted flow.

Node

A Node on a network is formed usually by the presence of a router and user access equipment - dial up, leased line or both. Often several leased lines are joined together at a network node. If a network topology is visualized as a road map, the leased lines are the roads and the nodes are the towns of which many roads travel. A POP is normally a network node, but a network node need not also be a POP.

Leaf

A leaf refers to a node found at the end of a network branch. There is only one connection between the leaf and the rest of the network.

Mesh

A mesh refers to one possible topology for a network backbone. For redundancy in the case of a circuit outage, a backbone is usually connected in a circular fashion so that if data can't get to the next node because of a line cut, it can get there by flowing in the opposite direction. A mesh architecture is formed by adding lines that go directly from one node to a second node three or four nodes around the "circle" from the first node.

Star

A star architecture would be formed by branches that would be connected to the network only at a central machine rather than connected to each other.

LAN

A Local Area Network (LAN) most often uses Ethernet as its protocol and generally extends through out a building or extends over several buildings over a radius of up to a couple of miles

MAN

A Metropolitan Area Network (MAN) refers to the high speed linking of hosts in buildings through out a city.

WAN

A Wide Area Network (WAN) refers to a network with a backbone that can link computers over distances of hundreds or even thousands of miles. T-3 or 45 megabits per-second has become a minimal WAN backbone capacity. However, some WANs still have only T-1 backbones.

CAP

A Competitive Access Provider is a venture that may compete with LECs and IXCs in offering a set of tailored data or voice services. A CAP will often be more regional that nation wide - although MFS is somewhat of an exception to this rule. A CAP is generally the equivalent of a CLEC.

Ethernet

Ethernet is a local area network transport protocol that first appeared in the 1970s. It offers a 10 megabit per second speed for data throughput. However, because hosts on an ethernet may transmit at random times, without any technology to deter data collision, the actual data throughput, depending on how heavily the ethernet is being used in a given situation, may be much closer to the three to six megabit per-second range. To complicate things still more 100 megabit per second Ethernet technology has been matured and gigabit per second Ethernet recently introduced. Analysts believe now that gigabit Ethernet will be increasingly used in pace of ATM in corporate LANs.

FDDI

FDDI or Fiber Distributed Data Interface is a 100 megabit per second transport protocol used over fiber in local area networks and based on token ring technology where data collisions are avoided by allowing hosts to transmit only when each host can grab the token or network equivalent of a green light for transmission.

Gigaswitch

A device used for high speed routing interchange at the MAEs to allow those with heaviest traffic to do their data exchanges before sending the remainder to the FDDI ring. At a NAP, when a FDDI ring becomes too crowded, the addition of a gigaswitch is generally necessary.

Service Quality

Not to be confused with Quality of Service, service quality refers to such standards among ISPs as 7 day a week by 24 hour a day staffing for network operations centers. Absence of busy signals for dial in customers. Availability of service without outtage. Help desk effectiveness. Small packet loss and so on.

RSVP

RSVP is the protocol used to establish a link between users wishing to reserve some specified amount of bandwidth. It was completed as an Internet Standard in September 1997. As of the end of 1997 no one has figured out how to charge for RSVP sessions traversing more than one network. This is another example of the intractable nature of settlements in the Internet.

Integrated Services

Integrated Services refers to an emerging set of applications designed to run on links established by the RSVP protocol.

Metropolitian Fiber Systems (MFS)

Metropolitan Fiber Systems is a major CAP that began in the late 80s by offering high speed Metropolitan Area Network service between clumps of strategically located buildings in roughly 20 cities across the US. During 1993 MFS established powerful backbone of its own. It then became a purveyor of bandwidth on a national scale when it used TCP/IP to link its MAN services together into a national Wide Area Network (WAN). It has since been acquired by WorldCom.

MFS

Metropolitan Fiber Systems is a major CAP that began in the late 80s by offering high speed Metropolitan Area Network service between clumps of strategically located buildings in roughly 20 cities across the US. During 1993 MFS established powerful backbone of its own. It then became a purveyor of bandwidth on a national scale when it used TCP/IP to link its MAN services together into a national Wide Area Network (WAN). It has since been acquired by WorldCom.

IP Switching

IP Switching refers to attempts to combine in a single device the benefits of switching at the IP level.

Internet Specific Data Terminology

IP number

An IP number (also referred to as Internet address number) in the Internet world is like a telephone number in the telephony world. IP numbers are the addresses of hosts or other intelligent devices on the Internet. The IP number of the desktop MAC/Internet host on which this glossary is written is 205.164.155.3

DNS

Domain Name Service offers a means of mapping a written name to an IP number. Thus one can write the easier to remember cookreport.com in place of the long IP number.

Top Level Domain

Also referred to as GTLD for Global or Generic Top Level Domain. Com (commercial) is a top level domain. Gov, net, edu, org (Government, network, education, organization) are the other top level domains that uniquely identify Internet addresses worldwide. Some domain name addressing is done on a geographic basis. For example cnri.reston.va.us refers to the Corporation for National Research Initiatives, Reston, Virginia, United States.

InterNIC

InterNIC stands for Internet Network Information Center. The most important part of the InterNIC is run by Network Solutions INC which is responsible for the assignment of domain names and IP numbers. On September 14, 1995 withe concurrence of the National Science Foundation, NSI announced a $50 charge per year per domain name.

Class C Network

A class C Network in the IP addressing scheme of things can accommodate 256 hosts. A Class A network holds 16 million hosts and a Class B network 65,000 hosts. Under IPv4 only 128 Class A address can exist - of which only 64 have been used. Class C addresses were to small for many organizations, which opted for Class B instead. When available Class B address began to be seriously depleted in the early 90s, CIDR (Classless Inter Domain Routing) was created to enable groups of class c address to be used together.

ASN

Autonomous System Number (ASN) Autonomous System has meant one of two things: (a) a set of systems sharing a common routing protocol under common administration, or (b) the domain of a routing protocol. RFC 1009 probably has the earliest formal definition: 1.1.3. Autonomous Systems -- For technical, managerial, and sometimes political reasons, the gateways of the Internet system are grouped into collections called "autonomous systems" [35]. The gateways included in a single autonomous system (AS) are expected to: Be under the control of a single operations and maintenance (O&M) organization; and Employ common routing protocols among themselves, to maintain their routing data-bases dynamically. A number of different dynamic routing protocols have been developed (see Section 4.1); the particular choice of routing protocol within a single AS is generically called an interior gateway protocol or IGP. An IP datagram may have to traverse the gateways of two or more ASs to reach its destination, and the ASs must provide each other with topology information to allow such forwarding. The Exterior Gateway Protocol (EGP) is used for this purpose, between gateways of different autonomous systems. [Our thanks to Fred Baker of Cisco for this material.] In short an autonomous system number (ASN) is the unique identifier for each autonomous system announced in the Routing Arbiter Database and in the route peering process in general. Internet Service Providers usually have one ASN. Roughly 7,000 ASNs have been assigned beginning with BBN which has ASN 1. Some NSPs have more than one ASN. But such use of multiple ASN's is declining.

CIDR

Classless Inter Domain Routing is described in a series of Internet RFC's (Requests for Comment). It was established in the 1992-93 time frame in order to allocate IP addresses more efficiently as the Internet began to grow dramatically. It is now the driving mechanism for the use of routing in the Internet as providers are straining to keep up with dramatic growth. Those wanting a detailed look at this extremely important subject should check the CIDR FAQ Sheet

Network Prefixes

Are used to aggregate class C networks. A network number with a prefix of 8 identifies a Class A, 16 a Class B, 24 a Class C, and 32 a single host. For example 207.8.160.0/19 would indicate the ability to serve 8,192 hosts beginning at address 207.8.160.0. As the prefix number grows smaller, the host ip numbers capable of being assigned from the CIDR block double with each decrease in prefix number. Thus 24, as a prefix, contains 256 host numbers, 23 holds 512, 22 holds 1024, 21 holds 2048, 20 holds 4096, 19 holds 8,192, and so on

Routing tables

Routing tables are lists of paths or routes to get to IP addresses that are held in router memory. A small router may have a table that takes an aggregated address (IE - this is for a host contained within the 207.8.160.0/19 hierarchy). Once sent to the gateway router, the routing table in that router will tell the router how to send it to its final destination via other routers lower in the hierarchy than the gateway router. Routing tables that must be carried by backbone routers have grown to exceed 30,000 routes. The most powerful backbone routers made by Cisco can carry about 60,000 routes. (And much more at the beginning of 1998.) Since the beginning of 1995 CIDR rules have been enforced much more vigorously both because the number of routes has been increasing faster than the growth of CPU and memory capability, and because a memory buffer is needed to allow routers to dampen route flaps.

Provider Based Routing

CIDR implementation has come to mean that routing has become largely provider based. An ISP get its IP numbers from its upstream provider's CIDR bloc and not from the InterNIC where only the larger providers can go. This operational necessity is causing power to gravitate into the hands of the larger players.

Geographic Based Routing

Dave Crocker, a prominent IETF member, has argued that the CIDR working group should have done more to enable geographic based routing where IP numbers would have been assigned according to geography rather than given by larger providers to smaller ones. Doing what he asks would be quite difficult. It would however lessen the ongoing concentration of power in the hands of the large providers.

Multiple Homing

When a small ISP begins to prosper, it often takes a leased line to a second (usually national) backbone provider. This is referred to as multiple homing. It requires the ISP to announce a second set of routes that cannot be CIDR aggregated and hence obviates the neat routing hierarchy that provider based routing tries to achieve.

Route Portability

Until 1995 if a customer of an Internet service provider wanted to move to another provider, that customer could take its IP numbers with it to the new provider. This however would "punch a hole" in the new provider's CIDR bloc and cause routes for the new customer to have to be announced all the way out to the backbone level. Consequently many providers began to say that they would not announce IP numbers that new customers brought with them. This meant that a customer would have to renumber, if it were going to change providers. Renumbering for a large customer is usually so expensive as to be prohibitive.

Route Aggregation

By assigning prefixes to old class C addresses and by apportioning their address (IP number) space carefully as well as encouraging customers to renumber and return unused addresses, ISPs can aggregate the announcement of many routes into the announcement of only one route.

Renumbering

In the future, renumbering of a customer's IP addresses will normally be imposed on a customer if that customer changes providers. If that customer is very large, the expense in changing IP numbers on dozens or even hundreds of hosts is likely to be vast. Also if the customer's applications are time critical, it will likely take the customer off the air for an unacceptable amount of time. Finally, some software applications actually require the IP number of the workstation on which they reside to be entered into the license. Change of IP number therefore could require a software license change.

Route Flap

When a network link (leased line) goes down, there will be routers which temporarily are unreachable by packets from the rest of the network. The network begins to route around the missing link and router. It does this by propagating from the point of the fault the information that the router is unreachable. Software in adjacent routers begins to redirect data and the need for data redirection is propagated, from router to router, throughout the network. This propagation spreads throughout Internet backbones like ripples from a rock cast into the middle of a pond eventually after a period of time reaching all the routers in the network. If however during this propagation time another link goes down else where, the changed routing data from that also propagates ripple like across the Internet pond. When the two sets of ripples intersect stress on router memory and CPU's becomes great and if network conditions are changing quickly enough the router software will exceed the ability of the hardware to cope. The router flaps or losses its ability to route. Such a flap can spread like falling dominoes across an entire network and, at a NAP, can threaten to cascade into the network of another provider. One reason for keeping the number of routes advertised well below the number the router can theoretically carry is that spare memory and CPU capacity are needed to enable a router to cope with chaotic conditions and avoid flapping.

NAP

On April 30 1995 the architecture of the American Internet underwent a major change from a single dominant NSFnet backbone to a series of commercial provider owned and operated coast-to-coast national backbones. Under these conditions, the backbones had to have some means of exchanging data. Four NAPs or Network Access Points were designated to serve as data interchange points for backbone service providers. NSF's three primary NAPS are in San Francisco California bay area, Chicago, and Pennsauken NJ near Philadelphia. Metropolitan Fiber Systems MAE-East was designated by NSF as a secondary Washington DC NAP. NAPs and MAEs are generally spoken of at the beginning of 1998 as public exchange points (IXPs).

MAE-East

Metropolitan Area Ethernet - East was built by Metropolitan Fiber Systems (MFS) as an interchange for PSI, UUNET and SprintLink in late 1993. Today MAE-East has nearly fifty attached service providers as users and was by far the busiest Internet hub in the world with data flow reaching in early 1997 more than 700 megabits per second when it was over taken in throughput by MAE West in the San Francisco area. MFS operates three or four other public exchanges known as MAEs.

Peering

Peering is the exchange of routing announcements between two service providers for the purpose of ensuring that traffic from the first can reach all customers of the second, and vis-versa. Peering takes place predominantly at NAPs and usually is offered without charge. One cannot however plug into a NAP and assume that those there will automatically grant peering.

Transit

Transit comes into play when a provider wants to reach customers of some third party that the first provider doesn't peer with. If the ISP that peers with the first provider also peers with the third party, then that provider is in a position to offer the first provider transit to the third party. Transit will normally cost a flat monthly charge.

Routing Arbiter

"Funded by the National Science Foundation, the Routing Arbiter project is charged with the task of providing routing coordination for the new NSFNET architecture. The project is a joint effort of Merit Network, Inc., the University of Southern California Information Sciences Institute, Cisco Systems, as a subcontractor to ISI, and the University of Michigan ROC, as a subcontractor to Merit." [Text from RA Web page at http://www.ra.net/routing.arbiter/RA/index.html.] If the routing Arbiter were working as intended, peering for many of those connected at the NAPs would be taking place via the route servers, relieving some of the stress on the current backbone< /A> routers. Because the Routing Arbiter Database was populated with material from the old NSFnet PRDB, some of the largest providers are refusing to use it. By the end of 1997 the Routing arbiter transitioned into a service made available by some of the public Exchange Points.

Route Server

"The RA project's Route Servers -- Sun SPARC 20 workstations installed at each interconnection point-- eliminate the need for pair-wise peering among the attached ISPs. The Route Servers input routing information from each ISP router, create a "view" (a Routing Information Base) reflecting that ISP's policy requirements, and pass the processed routing information to each ISP's router. T he Route Servers thus reduce the number of peering sessions each ISP router needs to process from O(n) to O(1). The Route Servers do not forward traffic at the interconnection points; they handle only the flow of routing information. For a technical description of Route Server services, see The RA Route Server Service Overview." [Text from RA Web page.]

Routing Arbiter Database

The RADB is "Successor to the PRDB, and one of several routing databases collectively known as the Internet Routing Registry (IRR). Routing policy is expressed in the RADB using RIPE-181 syntax. Analysis code developed jointly by ISI and Merit processes customer data entered in the RADB and produces GateD and Cisco router configuration files for the Route Servers." [Text from RA Web page.]

NANOG

The North American Network Operators Group is composed of Internet Service Providers who have technical and operations oriented meetings three times a year and who maintain a useful mail list (nanog@merit.edu) for the discussion of Internetwork operations issues. NANOG had itself grown so large by the end of 1997 that the leading technical people of the major providers complained of their inability to get serious technical work done.

Internet Service Provider Terminology

ISP

An Internet Service Provider is an entity that provides commercial access to the Internet. These can range in size from someone operating dial up access with a 56 kilobit line and several dozens of customers to providers with multiple pops in multiple cites and substantial backbones and thousands or even tens of thousands of customers.

NSP

A National Service Provider is an Internet service provider of national scope - one that would provide service in many different states and be connected to public exchange points on each coast. There is no clear dividing line between a small NSP and large ISP.

Regional Network

A Regional Network usually refers to one of the academic and research oriented nets started in the late 1980s with help from the NSF. There were about 30 of these and virtually all are still functioning. A Regional Network may now also refer to a large ISP that is not yet really national in its scope.

R&E

Research and Education (R&E) refers to the clientele served by the NSF's networking efforts. The old NSFnet acceptable use policy (AUP) divided traffic into R&E compliant and commercial or R&E non compliant.

POP

A Point of Presence (POP) refers to a node of an Internet service provider (ISP) containing a DSU-CSU, terminal server and router and sometimes one or more hosts, but no network information center (NIC) or network operations center (NOC).

DSU-CSU

A DSU-CSU is the digital equivalent of a modem. It is used for connecting a leased line, usually to a router. Ascend, Livingston and several other companies began offering units in 1995-96 that combined the function of digital modem, terminal server, and router into one compact unit. These devices have become quite standard among all but the tiniest ISPs.

NOC

A Network Operations Center (NOC) is the nerve center for an ISP where a seven day a week 24 hour a day staff is on duty to monitor equipment and correct problems. Equipment at POPs with no humans on duty can often be fixed remotely by some one at a NOC.

NIC

A Network Information Center (NIC) covers services like domain name service (DNS) and customer assistance. NIC functions are sometimes merged with NOC functions - that is to say NOC staffers also handle NIC kinds of duties.

SLIP

Serial Line Internet Protocol (SLIP) is software that an ISP customer can use with a modem to dial up a terminal server and make an IP connection to the network. The ISP must have a unique IP number to assign to each slip customer. Netscape and other web browsers will work only with SLIP or PPP connections to the Internet.

PPP

Point to Point Protocol is a variation of SLIP used for essential the same purpose.

Terminal Server

A terminal server is a device into which modems located at a POP are plugged. The terminal server attached to the router which in turn is attached to the DSU-CSU.

Modem

A modem takes digital data from a computer and converts it into electrical pulses that can be sent over a telephone line where they can be received by a second modem and converted back into digital data. State of the art modems currently can send 33,400 bits of data over an ordinary telephone line. It is unlikely that modem technology will result in a significant increase in this speed. So called 56,000 kbs modems appeared in mid 1997 but there is as yet no standard and these modems also do not function at anywhere near their optimum speed except in the most optimal local lo op conditions.

Server

A server is normally thought of as a powerful computer that can answer queries from clients. The client-server function is usually some variation of a database function. That is to say the client asks the server to send information to the client. The client software, usually residing on the machine of each end user, is tailored to work directly with the server software.

Web

The World Wide Web is software that enables digital data that has been "marked up" with HTML to be put into hypertext databases where data in one database can be linked to data in another so that by clicking on the marked text the user is automatically and transparently logged into the new web server where the linked data is found. In 1994 -1995 the World Wide Web became the driving force behind Internet growth.

Web Browser

A Web Browser is client server software used to query world wide web sites. Mosaic was the first popular browser. Netscape however has come to dominate the market with about an 80% share.

Hypertext links

Hypertext links make it possible to connect or link a passage in a document to another document on the same web site or a document on web site on a machine residing on the opposite side of the world. One click on the link takes the user automatically to the document to which the object that has been clicked on is linked.

HTML

Hypertext Mark Up Language (HTML) is the code that must be applied to data to produce desired displays of pages on the world wide web.

Usenet news

Usenet news is composed of more than 35,000 subject matter discussion groups where data is composed of ascii text and binary files. More than six gigabytes (billion bytes) of new data is generated every day. Access to usenet news is an expected part of Internet service.

File Transfer Protocol (FTP)

FTP is the application program used to send or receive large ascii or binary files over the Internet. More and more however such files are being sent and received by means of access to web sites.

Gopher

Gopher was the first software developed to make the use of FTP easier for the non technical user. Like FTP Gopher has been subsumed by the world wide web.

Telnet

Telnet is the application program used to reach the login prompt of any computer not behind a firewall on the Internet. The syntax is command followed by domain name as in: telnet tmn.com.

Host

In general terms a host is a single or multi user computer that can send and receive data over the Internet.

IPv4

IPv4 is the current version of the IP protocol in use today. Several aspects of the protocol today such as numbers of addresses available are not scaling very well in the current exponential growth of the Internet.

IPv6

IPv6 (formerly referred to as IP(ng) or IP next generation) is a new version of IP designed for the new commercial Internet. Test implementations are underway, but some people believe that the embedded nature of IPv4 is so vast that the costs of changing networks and applications to IPv6 will be huge and therefore will take many many years to bring about.

Internet Related Organizations and Associated Terminology

IAB

The Internet Architecture Board (Formerly Internet Activities Board) is the coordinating and oversight body for the actions of the Internet Engineering Task Force (IETF) and the Internet Research Task Force (IRTF). In June of 1992 the IAB, IETF, and IRTF were given a new legal home under the aegis of the Internet Society.

Internet Research Task Force (IRTF)

The IRTF develops and carries out Internetworking research experiments.

IETF

The Internet Engineering Task Force is the standards promulgating body of the Internet. It has a very successful record of developing standards such as the Simple Network Management Protocol that are quite quickly adopted by major segments of the network industry.

IESG

The Internet Engineering Steering Group (IESG) is the governing body for the IETF. The IESG issues last calls and gives final approval to new protocols.

Internet Secretariat

The Internet Secretariat provides logistical and administrative assistance to the various Internet governing bodies (IAB, IETF, IRTF, IESG, etc). Agencies belonging to the FNC have given enough funding to the Corporation for National Research Initiatives (CNRI) to allow it to house and pay for the expenses of the Secretariat which includes staff positions for Executive Director of the IETF and a Secretary for the IESG.

Requests for Comment

Requests for Comment (RFCs) are individually numbered official Internet documents that give information about Internet standards specifications, organizational notices; and individual points of view. Information on how to access RFC files may be found in most standard published technical guides to the Internet.

Internet Assigned Numbers Authority (IANA)

The Internet Assigned Numbers Authority was little known before the emergence of the DNS wars of 1996-97. Since it was created in the early 1980s it has been in the hands of Jon Postel, a highly respected creator of the ARPANET. Jon fulfills the role single handed - although he has the assistance of some staff help. He has the final decision making authority for the operation of DNS, for the Root Name Servers, for the assig nment of IP numbers, for the handling of appeals regarding IP numbers, for the RFC series, for the .us top level domain and for port number assignments for protocols. In 1995 the IAB and IESG agreed not to challenge IANA decisions. Jon Postel as IANA sits on the IAB. Many people now believe that the demands and size of the commercial internet have reached a critical point where a single person can no longer try to act as a "supreme court" for the Internet. These people say, justifiably we believe, that serious reforms of the IANA functions are needed immediately.

FARnet

The Federation of American Research Networks is an association in which the mid-level networks of the NSFnet, two commercial providers (ANS, and PSI), and some of the telephone companies meet usually four times a year to discuss common interests. With the complete commercialization of the Internet FARnet has declined considerably in importance.

FCCSET

The FNC reports to the Federal Coordinating Committee on Science Engineering and Technology. FCCSET in turn reports to the executive branch's OSTP (Office of Science and Technology Policy). FCCSET was required by the High Performance Computing and Communications legislation to provide a report to the Congress by December 1992 on the planned implementation of NREN. As of 1997 it no longer plays a noticeable role.

NREN

The National Research and Education Network (NREN) was Al Gore's dreamed of government funded, gigabit-per-second hi-tech backbone for the American Internet. It will never be built.

EDUCOM

A national lobbying organization composed primarily of major academic computing centers.

ISOC

The Internet Society was launched in 1992. It had been put together even earlier by a group of about a dozen long time and well respected Internet professionals. In 1992 CNRI, Educom, and the Association of European Research Networks (RARE- since renamed Terena after a merger with EARN) became founding members of ISOC with the basic privilege that ISOC by-laws could not be changed without their approval. This has lead to some discord with in ISOC where the group that wishes to keep it primarily a society for individual Internet engineering professionals is dominant over a minority that wishes to encourage industry wide support. The three founding members of ISOC no longer have veto power over by-law changes, however many members of the commercial Internet industry have reacted negatively to the role played by ISOC in support of the Internet Ad Hoc Committee (IAHC) on DNS. ISOC enjoys considerable support, but it can by no means be said to speak for the Internet, because, as of the beginning of 1998, neither it nor any single organization enjoys universal support of the numerous interests making up the commercial Internet. Some believe that the days when any single organization involved with the Internet can enjoy universal support are gone.

CIX

Commercial Internet Exchange was initially the agreement between PSI, Uunet, CERFnet, US Sprint (Sprintlink), that lets the traffic of any member of one network flow without restriction over the networks of the other members. Any TCP/IP service provider may join the CIX for a cost of $7,500 and connect to and send traffic to other CIX member networks. Until a major dispute over whether it would filter the routes of downstream customers of member networks who were not also CIX members at the end of 1994, the CIX represented the vast majority of the commercial Internet. With the opening of the NAPs and MAEs as public exchange points in 1995 the CIX's influence has greatly diminished. In 1997 the CIX is still about the size it was in 1995 and still operates its public exchange point which has been abandoned by almost all of the very largest providers.

FNC

The Federal Networking Council is the coordinating body for the networking interests of the so-called federal mission agencies. It is run primarily by representatives from DARPA, DOE, NAS, and NSF. Several other Federal Agencies are represented, but without significant power because relatively little money is authorized for them.

CNRI

The Corporation for National Research Initiatives (CNRI) was founded in the 1980s by Robert Kahn (co-author with Vint Cerf of the TCP/IP protocol) as a civilian Defense Advanced Research Projects Agency (DARPA). CNRI currently houses the IETF Secretariat and the Cross Industry Working Team (XIWT)

MILnet

Military Network is the production, non-classified TCP/IP network of the Defense Department.

Esnet

Energy Sciences Network is the TCP/IP network of the Department of Energy. With the rise of the commercial Internet, it has declined considerably in importance

NSInet

NASA Sciences Internet is the TCP/IP network of the National Aeronautics and Space Administration.

NSFnet

The National Science Foundation Network was expected to become the core network of the National Research and Education Network or NREN. The NSFnet was composed of the backbone of 19 sites or nodes and 32 mid-level or regional networks that connected more than 1000 institutions to the backbone. NSFnet, until its turn off on April 30 1995, was the major backbone of the Internet. The NSFnet regional networks now are connected to the three primary NAPs by either MCI or Sprint.

FIX West and FIX East

Federal Internet Exchange West is a link at NASA Ames (Moffit Field) California and Federal Internet Exchange East College Park Maryland between the backbones of MILnet, ESnet, and NSInet with the NSFnet. With the dissolution of the NSFnet, the FIXes now connect to the Internet at MAE East and MAE West.

MERIT

Michigan Education & Research Information Triad (MERIT ) was the holder of the 1987 cooperative agreement with the National Science Foundation for the provision of the T-1 and then the T-3 NSFnet backbone. The five year agreement lasted 7.5 years and ended on April 30 1995. MERIT maintained a subcontracting relationship with IBM and MCI as joint study partners and since September of 1990 with Advanced Network and Services. MERIT now was one of two awardees for the Routing Arbiter Cooperative Agreement. It is now involved in the Internet 2 project.

ANS

ANS Advanced Network & Services is the 501(c) (3), non profit, IBM and MCI spin off corporation launched on September 17, 1990. Launched with a $5 million dollar contribution from each of its corporate parents, it found itself in a position to inherit control of a privatized National Research and Education Network. From 1990 ANS received between $10 and 15 million per year from the National Science Foundation for providing the T-3 backbone to which it was also free to sell commercial access. The ANS/NSFnet backbone connected 32 mid-level networks which in turn connected over 1,000 institutions. ANS was purchased in late 1994 by America On Line for $35 million. In late 1997 it is to be acquired by WorldCom for $100 million,

American Registry for Internet Numbers (ARIN)

American Registry for Internet Numbers (ARIN) was created in 1997 to separate IP registration from the DNS service operated by InterNIC and owned by Network Solutions Inc. ARIN is a membership owned and operated entity designed to distribute IP numbers delegated to it by IANA and to set policy for distribution of IP numbers to end user ISPs in accordance with member wishes.

Internet Business Model Terminology

Settlements

Settlements comes from the old telephony terminology where if one party makes a disproportionate use of another's network, one has to pay that network monetary compensation. Settlements have been talked about for Internet service providers. The assumption has been that they might be based on numbers of gigabytes sent and received. This would likely lead to some form of measured usage charging - something that is anathema to Internet culture - where charges have always been based on the size the bit pipe leased by the provider. We believe that it is unlikely that settlements can be imposed in the absence of unanimity among the major service providers, for those who imposed settlements would likely loose customers to those who did not. Charging for route announcements, however, is really another form of settlements. As there is some necessity to keep the number of routes announced from growing as rapidly as it has been, there is a distinct possibility that settlements via route announcements might be imposed.

Guaranteed Service Standards

The lack of guaranteed service (now called Quality of Service) is a complaint of those who want to impose settlements - something that would intensify power in the hands of the large providers and be likely to drive many small providers out of business. Because of the disparity of resources between large and small providers, there is a considerable divergence in both Service Quality and Quality of Service. Yet the net works as well as it does because of cooperation between entities which also must compete with each other. One of the tensions most strongly felt in the commercial Internet is where the line between competition and cooperation should be drawn.

Metered Service

Metered Service is a short hand term for charging for measured use by amount of data sent and received. A point of confusion arises because some people also use metered service to refer to charging dial up customers a fee for the amount of time they remain connected to the network.

Route Announcements

All networks have to announce routes to their upstream providers in order to be reachable from the rest of the Internet. Constraints in the ability of routers to handle the explosive growth in numbers of routes announced have led to pressures to reduce these numbers. Discussion is beginning about imposing charges for each route announced as a means of further discouraging large numbers of route announcements. Imposing such charges would be tantamount to settlements in a different form. Depending on how it were done, it could likely be a significant force in driving smaller providers out of business

Stupid Network

The term "stupid network" was coined by David Isenberg in his famous June 1997 paper Rise of the Stupid Network: "Why the Intelligent Network was once a good idea, but isn't anymore. One telephone company nerd's odd perspective on the changing value proposition." Isenberg's concepts (they aren't really his - he just articulated them more forcefully and clearly than anyone else) are absolutely critical to a understanding of the business model of the next gen telcos and the wave of optical networking that will now begin to sweep over the telecommunications industry. Fundamentally, the argument is that when bandwidth and memory were expensive and scarce, the telcos built networks with intelligent, multi-million dollar switches at the core. Thus the "intelligent network" was the most efficient way to provide service. Now we have new developments in fiber (WDM, etc) and the IP protocol as a common network communications language. With bandwidth ever more available and memory available at a few dollars a megabyte, the core intelligence of and control over the old telco networks is migrating with great speed to the desktop of the network user. We are talking about a revolution more profound than that of the PC versus the mainframe in computing. Telecom will never be the same. In looking at a design principal or technology, readers need to internalize these concepts well enough to be able to answer whether that principal or technology reflects its creator's understanding of the "stupid network".