6.033 Assignment: Recitation 13, 2004-03-18

M.I.T. DEPARTMENT OF EECS

6.033 - Computer System Engineering

Recitation 13 - Thursday, March 18, 2004

Network Address Translator (NAT)

Read the paper The IP Network Address Translator (NAT) by Egevang and Francis. This paper is an RFC. In contrast to most of the other papers your read in 6.033, RFC's are not refereed papers. They are technical documents generated by the Internet Engineering Task Force (IETF) to describe the details of a networked system and facilitate its implementation by interested developers. Usually they do not contain any performance results. If you would likek to know more about RFC's, see the RFC overview page.

Also read Things that NAT's break. The origins of this document are slightly unclear, but we believe that the document was compiled by Keith Moore (U. Tennesee), who may also be the author of the document. N. B.: Neither of these papers is in your reading package; they are available only on-line.

NAT tries to address the shortage of IP addresses by allowing address re-use. The RFC includes references to CIDR, another solution to address the shortage. It includes references to many other acronyms. You don't need to understand these acronyms to understand the NAT paper, but in case you are curious you can read:

NAT, which is discussed in the paper, advocates address reuse. Read the abstract and sections 1 and 2. Look carefully at figure 2. Make sure that you understand how the sender addresses its receiver, and how the IP addresses in the packet change as it crosses the NAT boxes.

While reading the paper try to answer the following questions:

Recall from lecture that to increase routing scalability, IP addresses are hierarchically allocated (i.e., the internet uses variable-length "area identifiers"). How does the hierarchical allocation affect the address shortage problem? Is the IP address space efficiently used?
What are the characteristics of a global address? What are the characteristics of a local/private address?
Why do global addresses have to be different from local addresses? What happens if a NAT box advertises its local IP addresses to the rest of the Internet?
Some FTP commands use IP addresses in its (TCP) payload. The paper states that when such a packet traverses a NAT box, NAT has to replace these IP addresses by a global address. Is this behavior consistent with the end-to-end argument? Is it consistent with layering? What are the risks of violating the layering?

Go to 6.033 Home Page