Why does the Princeton CS department have a NAT box if we have enough IP addresses? Security - this way, the actual addresses in use for most of the machines don't get sent. So, for example, if a web site is trying to track accesses to individual machines, they'll have a hard time trying to gain any info from the IP address alone. What was the business about running NICs in promiscuous mode on subnets? If you're using subnets, then you need to send data to a router to be sent to machines on another subnet. However, it may be the case that two sets of machines share the same physical network but have different subnet info, in order to "isolate" their traffic. That's fine for many purposes, but if you really care about security, this is not sufficient, since any machine can put its own network interface into promiscuous mode and see whatever traffic is on that segment. I should mention, however, that with switched networks, all you're likely to see is the broadcast traffic. The switch won't send other traffic to machines if those machines aren't involved. If you're connected via hub instead of a switch, then you will see the traffic. Why does an AS need to store information about nodes not strictly above or below it? (stems from the sideways BGP diagram in the lecture) Note that BGP assume arbitrary topologies, not just trees. The diagram in the slides (and the book) showing the various AS's is just a simplification. The reason for storing entire paths is to allow finding a workable path and ensuring it is loop-free. Does the subnet mask always cover the network part of the IP address? Yes - remember that you're basically dividing up your allocation into a group of smaller networks, so you'll always be including your network number in the subnet mask. What happens when different routers use different gateway protocols, or is this not allowed to happen? From reading the following http://www.cisco.com/warp/public/cc/pd/rt/2600/prodlit/2600_ds.htm It appears that the common Cisco routers support BGP and OSPF. That would be enough for them to act as border routers in the internet, as well as handling traffic within their own AS. Couldn't you use CIDR to address Japan's IP shortage problem, by growing the number/range assigned to them? Nobody's keen on giving up ranges given to them, even if they're not using all of them. I believe that there used to be draft RFCs that were proposed asking IP hogs to give up some of their addresses. What is holding IPv6 back from greater deployment? Primarily just the existence of CIDR and NAT? Is IPv6 ever going to happen, and if so, when? See the dancing turtle - doesn't it make you want to upgrade? http://www.kame.net/kame-mosaic.html Yes, CIDR and NAT are responsible for getting rid of a lot of the motivation for IPv6. Realistically, the impact of switching to IPv6 isn't just about changing a few addresses, but also changing all applications that are expecting a.b.c.d format IPv4 addresses. So, unless there's some really big motivation or money to be made, why bother? The things that might cause a growth in the need for visible IP addresses are true peer-to-peer systems, lots of mobile devices, etc., etc. To understand the power of Japanese marketing, consider that Domo-Kun (http://www.cardhouse.com/japan/domokun.htm) has a 90% recognition rate among Japanese children, and that Pokemon is known world-wide. If Japanese technologists cooperated with the marketing folks, the world could be forced to use IPv6 :-) On slide 23, can you go over the stuff about BGP not needing unique AS numbers on leaf ASs? See page 323 of the book. Basically, AS2 and AS3 are advertising the fact that the various networks ranges in ASs4-7 can be reached via them. So, since there's someone advertising the addresses, and they're only one AS hop away, there's no need to have a unique AS number for the stub ASs. Is there a relationship between AS numbers and IP ranges, and if so, what is it and how is it expressed? There's no fixed relationship. The rules are basically that an AS is responsible for routing within the AS, and that's how BGP scales. So, if an AS has computers spread out over the network and wants to advertise itself as a single AS, then it has to have its own way of routing within the AS, presumably using leased lines, etc. Note that an AS can be reachable via multiple other ASs - it's a multihomed AS instead of a stub AS. What is an edge router, and what is the difference between an edge router and a border router? I believe that border router and edge router are used more or less interchangeably. Page 320 of the book defines border router as the router that sits between the AS and the rest of the internet. If a small company had a connection via an ISP and had their own router, that router would no doubt be an edge router. Whether or not it's a border router is probably a matter of taste. How long does it take for the Internet to recover from a major router or link going down? I personally don't know the answer to this, but if I had to venture a guess, I'd say that it's on the order of 30 minutes or less. That's mostly based on networking people telling me that routes stabilized very quickly after 9/11, even though the infrastructure was badly damaged around New York and though many sites were overloaded from traffic. Does FEC allow Company A owns 172.200.x.x/16 and Company B owns 172.200.3.x/8 (presumably /24) If not, then I'm confused Take a look at the bottom of page 317 where it discusses route aggregation. Now, imagine that one of the companies is multi-homed and being advertised via a different router as well. So, the router doing the aggregation would have a short prefix, while the other router might see as well. If the multi-homed company does not carry any transit traffic (and we wouldn't expect that it would), then its router would have both the more specific route for it as well as the aggregated route to the other router. These readings take a long time Yup. There's not a whole lot I can do about that, since I'm trying to follow last year's format to the extent possible. However, the worst that happens if you don't manage to do all of the reading on time is that you might get lost in lecture. Just do it before the midterm and you should be fine. You owe Wilkie money. You also need your telescoping pen. I should have both in class. Someone remind me.