[ Pobierz całość w formacie PDF ]
.Suppose the internal address was10.7.54.192 and the NAT box mapped that to the global address 105.197.183.114.Thatmeans the message would be 4 bytes bigger after the address in FTP was replaced.Theproblem is, TCP numbers bytes, so now the NAT/NAPT box has to also remember that it hasto adjust the TCP sequence number on every subsequent packet of that FTP connection.That would be particularly difficult to coordinate across NAT/NAPT boxes.End-to-end encryption would make it impossible for the NAT/NAPT box to look inside andreplace addresses.Another idea being introduced is called host NAT or host NAPT.The idea is that the hostinside must first request a global address from the NAT/NAPT box before initiating aconnection, but it is the host that does all the necessary translation of the packet.I think this isa more promising approach.Homework1.With IP addressing (and noncontiguous subnet masks), find two different (value,mask) pairs and an IP address that would match either link.Make sure that eachmask has at least one 1 where the other mask has a 0.2.(research) Suppose that a router had a list of destination links, as a table of (value,mask) pairs, with instructions for each entry about how to route a packet given adestination address matching that entry.Assuming noncontiguous subnet masks, isthere, in the general case, any algorithm (including any way of organizing the table),other than trying each (value, mask) pair in turn, that will indicate which entrymatches the destination address? (I think the answer is no.It would be of greatinterest if someone were to devise an efficient algorithm.)3.Assuming that noncontiguous masks in IP are disallowed, IP routing becomes similarto ISO level 2 routing the (value, mask) pairs become address prefixes and can beexpressed as a (mask length, value) pair.How many bits must be allocated to the mask length? Is an encoding of (mask length,value) more efficient than an encoding of (mask, value)?4.Some people have attempted to get noncontiguous IP subnet masks banned, butthere are those who defend noncontiguous subnet masks as being useful.Can youthink of anything that can be done with noncontiguous subnet masks that cannot bedone without them?5.Assume a problem that requires finding, from a table of entries, the entry that is thelongest matching initial substring of a given quantity.Devise an algorithm to do thisefficiently.6.Suppose you are told that you must implement a high-performance ISO level 1 routerand that the customer must be able to set a management parameter indicating whichsize ID field is to be used.What kind of strategies could be employed? Keep in mindthat high-performance forwarding usually requires specialized hardware.What kindsof economic trade-offs can be made? For example, the router might theoreticallysupport all address sizes but attain higher performance on some.Also keep in mindthat your product should be competitive both economically and in terms ofperformance.7.The rule in IP is that none of the three fields (net, subnet, host) in an IP address isallowed to be all 1's or all 0's.Why does this rule exclude a single-bit host or subnetfield?8.At $1,250 per OUI, what percentage of Bill Gates's net worth would be consumed bybuying up all OUIs?9.Compare the addressing schemes of IPv4, IPv6, CLNP, and IPX.Consider ability toautoconfigure, bandwidth use, simplicity, interconnection over public networks, andanything else of importance.Chapter 10.Connectionless Data PacketFormatsThis chapter covers the data packet formats of various connectionless network layerprotocols, including IP, CLNP, IPX, DECnet, AppleTalk, and IPv6.Again, you might think thatall you care about is IP, but it's interesting to compare approaches.Also, you'll understand IPat a deeper level if you compare it to other things, and there is no other reference for seeingall the packet formats in one place.10.1 Pieces of a Connectionless Network LayerConnectionless network layer protocols tend to be described in several documents.The moreor less separable pieces consist of the following.1.Basic connectionless service:This part of the network layer consists of the format for data packets and certain errormessages or other notifications that the network can send to an endnode.In CLNP (ConnectionLess Network Protocol), this part of the protocol is defined inISO 8473, "Protocol for Providing the Connectionless-mode Network Service." I referto the ISO protocol by its common nickname, CLNP.In the Internet protocol suite theequivalent protocol is defined in RFC 791, known as "The Internet Protocol," or IP.IPv6 is defined in RFC 1883.I found the specification for DECnet Phase IV at thefollowing URL:http://ftp.digital.com/pub/DEC/DECnet/PhaseIV/route20.txtAppleTalk is nicely documented in the book Inside AppleTalk, by Gursharan S.Sidhu,Richard F.Andrews, Alan B.Oppenheimer, and Apple Computer, published byAddison-Wesley.2.Neighbor greeting:This aspect of the protocol enables neighbors to discover each other.It allowsendnodes to know which routers are available on their LAN and to find out that it ispossible to communicate directly with other endnodes on the same LAN.It alsoallows routers to find their endnode neighbors.In this way, the routers can tell all theother routers how to reach those endnodes, and the neighbor routers know the datalink layer destination addresses needed to deliver packets across the final hop todestination endnodes.I talk about this part of the network layer in Chapter 11.3 [ Pobierz całość w formacie PDF ]