Networking I 
Fall 1997 
Questions: Stevens, Chapter 1


1.  From the bottom up, what are the layers of the TCP/IP protocol suite?

link, network, transport, application

2.  What is the difference between UDP and TCP transport services?

The User Datagram Protocol (UDP) differs from the Transport Control Protocol 
(TCP) in that UDP does not guarantee delivery (TCP does).

3.  What are the differences between network layer protocols and transport
    layer protocols?

Network layer protocols address nodes on a network.  Transport protocols are 
able to address specific process on a node.  Network layer protocols send
data to directly connected nodes, while transport protocols can send data
across the entire network.

4.  How has the meaning of the term "gateway" changed from its original
    usage to its usage today?

Gateway used to refer to a router:  a machine which connected two networks.
Today, gateway often refers to an application gateway:  a process which
connect two different protocol suites.

5.  Explain the relationship between a multihomed system and a router.

Any system with two or more network interfaces is called multihomed.  Unless 
the system forwards data between the interfaces, it cannot be called a
router.

6.  What is the function of "port numbers?"

These allow transport protocols (such as TCP and UDP) to identify specific
processes running on a host.

7.  Explain the differences between iterative and concurrent server types.

An iterative server handles only one client at a time.  A concurrent server
spawns a new server for each client request.  If the system is
multiprocessing, this permits multiple clients to be serviced at once.

8.  Name the two common variants of TCP/IP protocol suite API's.

Berkeley sockets and TLI are two common API variants.  XTI is a superset of
the TLI API.

Questions: Stevens, Chapter 2


1.  Describe how ethernet and IEEE 802.3 encapsulation differ.

In an ethernet frame, the data immediately follows the destination, source,
and type fields.  In 802.3, 3 bytes of 802.2 LLC and 5 bytes of 802.2 SNAP
follow the destination, source, and length fields (the type field is
included in the SNAP, just before the data).

2.  What base-10 number does 0xc0 represent?  What special character does it
    specify in the SLIP protocol?

Decimal 192 (hex 0xc0) represents the END character in SLIP.

3.  Name the three deficiencies of the SLIP protocol mentioned in this book.

Each end of the connection must know the other's IP address.  There is no
type field, so if SLIP is used over a serial link, no other protocol may be
used at the same time.  There is no checksum data, so if errors occur,
another protocol layer must detect the problem.

4.  How does PPP differ from SLIP?  Which is the preferred protocol for
    normal usage?

PPP permits the use of multiple protocols over the same serial link.  There
are cyclic redundancy checks on every frame.  There is a provision for
dynamic negotiation of IP addresses for each end.  There is a link control
protocol for negotiating many data-link options.  PPP is the preferred
protocol for serial links (if it is available).

5.  What is the purpose of the loopback interface?

It permits a client and a host on the same computer to communicate using
TCP/IP.

6.  Can data sent to the loopback interface be sniffed by another host on
    the same ethernet?  Explain your answer?

No.  Data sent to the loopback interface gets immediately copied to the
input queue, not sent out on the network.

7.  What is the largest possible MTU over Ethernet?

1500

8.  Why would the larget possible MTU differ between an Ethernet link and a
    PPP link?

Since serial links are most often much slower than ethernet links, it may be 
necessary to adjust the MTU on the serial link to avoid unacceptable
interactive response times.

9.  Describe symmetric routing.

The route from A to B is the reverse of the route from B to A.

10. What mechanism does TCP use to set the MTU for a transmission?

TCP uses the path MTU discovery mechanism.