LIS353 Spring 2001 Mid-Semester test

Answer questions 1, 2, and 3. Answer any two of 4, 5, and 6. You may
use books, articles, notes, and computers to complete the questions,
but you may not solicit or receive assistance from other human
beings. You may turn in the exam as hard copy, or email it to me. Any
diagrams should be turned in as either hard copy, encapsulated
PostScript, PDF, JPEG, or PNG. Make sure your name is in the content
of any files you mail me (not just in the mail header).

QUESTION 1: The protocols we're examining have emerged from two
different standardization enterprises. Some -- Ethernet for example --
are ISO standards developed by groups like IEEE (or ANSI, NISO, etc.)
But the core protocols described in the Hall text come out of the
IETF's Internet standardization process. Both approaches result in
layered protocols that employ encapsulation of data. But ISO's OSI
reference model has a somewhat different conception of what the layers
are and where they divide. Is this a problem for applications that use
both ISO and IETF protocols?

Take a position that the differences between IETF and ISO
standardization (conceptual, procedural, and/or technical) either are
or are not problematic for effective cooperation among these protocols
in the same systems. Back your argument with as many specific details
as you can of the protocols themselves, the reference models in which
they're situated, and/or the standardization processes.

QUESTION 2: Suppose I plan to develop an application layer protocol
for emulating live audio conference calls for groups of 2-10 people
over the Internet. The participants will speak into headsets that are
connected to the sound cards installed in their PCs.

Should I definitely use TCP for session and transport services, or
is UDP a better choice? Would it depend on other factors that I
haven't specified? Take a position on this question, and back
your argument using specific details of the TCP and UDP protocols.
The appropriate level of specificity is that adopted by Hall in our
text, but you're welcome to refer to other resources in addition to
the Hall book.

QUESTION 3: The IP header includes Type-of-Service flags that allow
one to request special processing of an IP datagram (e.g., minimize
delay, or maximize reliability). But ISPs don't generally treat
TOS-marked datagrams any differently than unmarked datagrams. Would it
be feasible for administrators of especially fast, secure, or
high-capacity WANs to charge money to the originators of TOS-marked
datagrams for the use of their networks? How would that work? Discuss,
if you can, other kinds of incentives -- either to ISPs, customers, or
both -- that would encourage the use of TOS-aware routers, but would
discourage users from arbitrarily marking all datagrams for special
handling.

QUESTION 4: At one point Hall suggests that a TCP segment could be
embedded directly into an Ethernet frame, rather than into an IP
datagram (page 269). If an application used TCP and Ethernet this way
(bypassing IP) would that eliminate the need for ARP, or would some
protocol functionally analogous to ARP still be necessary? Explain why
or why not.

QUESTION 5: Suppose I learn that a streaming video broadcast from the
US Senate is being sent to a particular multicast group address.
By pinging an arbitrary multicast address I learn that my local router
is a multicast router. Can I now conclude that I should be able to
access that Senate video by sending an appropriate IGMP message to
my local router? Explain why or why not.

QUESTION 6: It's been suggested that some firewalls refuse to pass
ICMP packets in the interests of security. Suppose you could program
a firewall to permit some ICMP messages and refuse others based on
the message type field in the header. Which messages would you want to
refuse in the interests of security and which would you wish to permit?
Provide a rationale for each recommendation.