Network software protocol suites supported in Earth Sciences
Last revision July 28, 2004
| Table of Contents: |
TCP/IP
The primary networking software protocol suite is TCP/IP, which stands for "Transmission Control Protocol/Internet Protocol". The "IP" part implements the lower levels of the OSI model to provide the addressing and data delivery functions. The "TCP" part implements the higher levels to manage actual connections and sessions. Although applications are not strictly part of the TCP/IP protocol suite, a set of basic application services developed on top of TCP/IP, such as remote login and file transfer, are normally included in any TCP/IP implementation and considered part of the protocol.
TCP/IP is the standard networking protocol for the world-wide Internet. All standard Internet services, such as email and web browsing, use TCP/IP. Organizations may use other networking protocols such as AppleTalk or NetWare on their local networks, but these will not be routed across the Internet. Thus, services provided by those other protocols will not be visible outside the local organization. TCP/IP is the native protocol for the Stanford campus network.
TCP/IP uses fixed names and addresses for computers, which requires configuration by network managers. You can't simply plug in a computer and expect it to work on the TCP/IP Internet unless it has been registered with local network authorities and properly configured. The advantage of these fixed addresses is that each computer has a unique name and address on the entire Internet, and therefore can connect to every other computer on the Internet.
The IP part of the TCP/IP protocol suite implements a hierarchical network addressing system. IP addresses themselves are numeric: 32 bits, divided into four eight bit integer values (each value ranges from 0 to 255) and normally represented in this format: nn.nn.nn.nn. For example, the IP address for pangea is 171.64.168.31.
The IP address is subdivided into "network" and "host" portions so a computer can tell from the address if the recipient can be reached directly on the same network or if the computer must send its data through a "router" that connects to another network. Routers can consult master servers maintained by the internet authorities to find out how to reach other networks.
People use names for computers when trying to use services, and generally don't like to use those IP address numbers. So there is a need to convert between names and IP addresses. This also provides a mechanism to organize groups of computers in a hierarchy called the "domain name system".
Each "fully qualified" IP computer name is actually composed of several parts. The first part is the individual computer "hostname" as known on its local network. The next part is a "domain" that organizes computers at the lowest level - perhaps by local network, or perhaps by a larger organizational unit. The next part is a higher level domain that indicates a broader organizational unit, and so on. There can be an arbitrary number of domain levels, but typically only two or three are used. The fully qualified IP name of a computer consists of all these parts, written together with the period or dot (.) character separating the parts. This full name is unique within the entire Internet. This full name can always be used when making network connections. When connecting to other computers that are located in the same domains as your own, you can normally use just the "hostname" part of the full name as a shortcut.
The last part of a IP hostname is called the "top-level" domain. It is normally either a two letter country code, or in the United States, it is typically one of the generic codes such as edu for institutions of higher education, gov for federal government departments, mil for U.S. military groups, and com for commercial organizations (and a catch-all for other types).
These top-level domains are controlled by the overall Internet naming authorities. They can allocate sub-domains (domain names at the next lower level) and blocks of IP addresses to other organizations which then control the allocation of names, IP addresses, and sub-domains within that. This hierarchical organization distributes the problem of maintaining the addressing information that connects IP hostnames to IP numeric addresses. Each domain (or sub-domain) authority sets its own rules for allocating hostnames (within its domain) and IP addresses to computers within its organization. Each authority then maintains a "name server" that provides name to IP address translations for its domain. At the top level, there are "well-known" name servers that each domain can contact to find the name servers and routers for other domains.
Stanford University has chosen to include most of the campus (with rare exception) into a single Stanford subdomain within the top-level EDU domain. Departments are not allowed to create further sub-domains within Stanford.EDU. This keeps computer addressing simple throughout the campus. You can use just the individual computer hostname part when trying to make a network connection on campus. It also means that new computers must be assigned a hostname that is unique within the entire campus. A campus-wide network registration system is maintained for this purpose.
AppleTalk
AppleTalk is a networking protocol suite that implements the complete OSI model, like TCP/IP. It was originally developed by Apple Computer Corporation to connect its Macintosh computers to each other and to LaserWriter printers using only LocalTalk (or PhoneNet) cables. It was then extended to work on ethernet (sometimes referred to as "EtherTalk").
AppleTalk support is built-in to every Macintosh, but not normally found on other types of computers. While still provided reluctantly in the new MacOS X operating system, Apple Computer plans to eventually phase out the use of AppleTalk. Third party software can be installed on Windows 95/98/NT/2000 computers and some Unix systems to let them communicate using the AppleTalk protocols. The reason to do this would be to let those other computer types provide or connect to Macintosh type services and AppleTalk printers.
The Windows NT and 2000 operating systems include enough of the AppleTalk protocol suite to allow them to access network printers that use AppleTalk or to provide file sharing via AppleTalk to Macintosh computers (for the server versions).
For Unix computers, one of the public-domain free programs Columbia AppleTalk Package (called CAP) or netatalk can be installed to provide AppleTalk connectivity. In Earth Sciences, we have installed CAP on the pangea server to provide file sharing to Macintosh computers via AppleTalk, and printer queues to access AppleTalk network printers.
At the application level, AppleTalk includes standard protocols for remote printing and file sharing services. There is no standard "remote login" as the Macintosh is not intended as a multi-user system, but "remote control" software using AppleTalk (or TCP/IP) can be purchased (for example, Timbuktu).
The AppleTalk protocol uses dynamic addressing. Routers that connect separate AppleTalk networks must use fixed addresses that are configured by the network manager. Individual computers and printers broadcast on the local network to find out about routers and select an unused address. Broadcasts are also used by computers to find each other's addresses and the services that are offered.
As there is no master authority controlling the assignment of addresses and names, AppleTalk cannot be used to create a large scale internet spanning across organizations. Stanford is able to maintain a single AppleTalk network for the entire campus, but AppleTalk cannot be routed beyond the campus.
In September, 2003, the campus Networking group plans to cease routing AppleTalk across the campus backbone network. At that time, any devices still using AppleTalk will be able to connect only to other devices on their only local subnetwork (the three Earth Sciences buildings in our case).