--Lisa Yeung:

Switches are capable of generating less network traffic by?
By selectively delivering messages to only the computers that need it.  Hubs
deliver (or broadcast) messages to every connected computer every time. On a
busy home network, one that utilizes Internet connection sharing and
high-traffic applications like music downloads and/or online gaming,
switches
handle this traffic noticeably better than hubs.

Which of the following technologies manage the conversion between IP
addresses and host names?
Domain Name System (DNS), Windows Internet Naming Service (WINS), and
Network Information System (NIS) each support a form of name-adddress
mapping on computer networks. WINS was designed for Microsoft Windows
networks and NIS was designed for Unix networks. DNS supports overall name
resolution and mapping on the Internet.

Which utility program is designed to report whether a networked computer is
responding at a given IP address, and how long these responses take?
The ping utility sends network packets using a special protocol to determine
the responsiveness of remote addresses.

The maximum size of an IP datagram (including the header) is?
The datagram length field in an IP header is 16 bits in length. So, the
maximum datagram size an IP datagram can support is 2^16 - 1 = 65,535 bytes



--Cal Woodruff

Q link/network layer: given that mac addresses are controlled why do we need
IP addresses?

Sample answer:
There are a number of issues that aren't addressed by mac addresses. 
One is that
many computers are mobile but routing, by design is regional in nature. You
would definitely not want the same identifier in a foreign country as you would
have in your home country.

Another issue with mac addresses is that you can only have one mac address per
network interface. Whereas there are many cases where it would be
desirable for a single
network interface have more than one ip address (eg load balancing).



Q transport layer: how would you characterize a tcp transmission capable of
reproducing human speech in a telephone conversation?

Questions:
Name 2 advantages of TCP for transporting sound packets vs UDP?

Sample answer:
One advantage is the sound samples are guaranteed to be sent in order. Ie the
sound won't be garbled the way it could be with a UDP transmission.

UDP sessions cannot tell when a call has been dropped. With TCP this will be
obvious.
----------------------------------
Why would a TCP and UDP "phone call" likely have equivalent performance
characteristics in practice?

Because conversations are full duplex, acks will always be accompanied
by meaningful data thus will be essentially equivalent to a udp session in
practice.
----------------------------------
Delay is an issue with telephone speech transmission particularly if packets
get lost. If packets get lost with a probability p what is the probability
of a perceived delay of d seconds? Assume a negligible decoding delay to turn
the packets back into sound.

Given n packets per second the "time slice" per packet will be 1/n seconds.

If RTT is smaller than d the number of lost packets needed to produce an
audible delay d would be d/n. The probability that this would happen would be
p^{d/n}.

If RTT is larger than d even one lost packet will create a noticeable delay
as packets won't be released from the receive buffer until the entire set of
packets are received. In that case the probablility would be p.



--Halim Soetanto

How do you impersonate someone's email with SMTP (spoof an email)?
- telnet to some mailserver with port 25 and then fill in everything as 
usual (HELO, MAIL FROM, RCPT TO, DATA, QUIT) but change the MAIL FROM 
clause to the person you want to impersonate as

How do you know if an email might be a fake (spoof)?
- one way is to take a look at the header of an email. If there's some 
inconsistency in the header (e.g. the email is coming from a 
mail server which is not the same as the mail server of the email address, 
it might be a hint that it's a fake!)




--Michael Jia

True or False?
1. Preamble is used to synchronize receiver, sender clock rates. (True)
2. Link layer has responsibility of transferring data from one node to the 
dest node via appropreiate path. (False)

Short Question:
1. Name 3 attributes an ideal multiple access protocol should have.
Answer: Efficiency (One node sends transmit at rate R, M nodes at average 
rate R/M), Fully decentralized, Simple

2. Compare ARP and link-layer switching carefully.
(1) What kind of mapping is kept in ARP table?
(2) What kind of mapping is kept in the link-layer Switch?
(3) How does ARP obtain the mappings in the table?
(4) How does the link-layer switch obtain the mappings?
Answer:
(1) IP to MAC address
(2) MAC address to interface
(3) Query and reply
(4) Self-learning from the source address of the data received 
from a particular
interface



--Mohamed Khalil

Provide at least 3 advantages that switches have over hubs.

1) Switch preserves isolated collision domains for each LAN segment 
connected to its interfaces.
2) Switch can interconnect different LAN technologies 
(10BaseT, 100BaseT, Gigabit Ethernet, กฤ)
3) No limit on how large a LAN can be when switches interconnect the 
LAN segments.
4) Switches can operate in full duplex.
5) Switches can provide cut-through switching.



--Alex Hunt


Name the transport and application layer protocols necessary for a client to
retrieve a document containing a JPEG image from an specific URL (where the
HTTP server is not know in advance).

Soln:

DNS for name resolution, UDP for the DNS, HTTP for webpage and JPEGs and TCP
for the HTTP.


Another might be:

As we saw in the second programming assingment that it is possible to
implement reliable data transfer over UDP. Give reasons why someone might
wish to do this:

Soln:
UDP is a connectionless protocol so there would be no setup overhead. Also
UDP  has not congestion control and would therefore not reduce its send rate
when the network became congested.