Is it me you're looking for?

LuciferDeathbringer I can see it in your eyes, I can see it in your smile.

***** You're all I've ever wanted, And my arms are open wide

whoops-- i thought i mentioned BGP as a path vector protocol. in fact, there are two link-state protocols in common use that i know of: OSPF and IS-IS. (i think i did say that, but it hit the cutting room floor :) EIGRP and RIP are distance vector rather than path vector protocols. (we could've gone into that as well, but it already seemed plenty long enough!)

Richard Mortier Dammit you're right, distance vector, not path vector. And does anybody use IS-IS? I've only encountered OSPF, perhaps because IS-IS is a layer 2 protocol and there is a move towards layer 3 routing in the distribution layer switches. (Or maybe because I never see the ISP side of things). Anyway, nice walk through, though I think I'd have been lost without a CCNA, CCNP background.

trefod thanks! back in 2001/2 when i spent 6 months at Sprint's IP research group, i wrote code to monitor their IP backbone routing protocols (github.com/mor1/pyrt) -- they used IS-IS then and i believe they still do (though no doubt someone will correct me if i'm wrong :) it was pretty interesting -- though IS-IS was updated to carry IP prefix information (RFC1195), it still runs over CLNP so i got to see bits of both. (if you're really interested, we used the data we collected to analyse the reliability characteristics of their backbone -- see www.cs.nott.ac.uk/~rmm/papers/pdf/imw02-linkfailures.pdf for the results :)

trefod Yes, some organisations use IS-IS - it's the routing protocol of the OSI model (most of the world and internet follows the TCP/IP model).

Dunno if you got a reply to this, but: There have definitely been cases of nation state-level actors causing "funny" routing to happen so they can conduct surveillance on a target, and also of criminal organizations rerouting traffic for dodgy purposes. I don't have any citations handy, but it shouldn't be too difficult to rustle some up with some Googling.

>want more visits to your store >do BGP attack to reroute everyone to your store >realise you just DDoSed your store

Egypt did it during the Arab Spring. They forced providers to reconfigure BGP. Much of the country could not interact on the internet.

+EgoShredder It is actually pronounced either way in America.

The way I understand it in the UK is a "rowt" vowel sound is used when something is drilled like a hole using a "rowter" (router). Things like network connections that define the route use the "root" vowel sound, because my network router does not drill anything. So that is the simple way of differentiating between drilling a hole or driving along a road to somewhere.

0 indexing is as God intended.

I'm not sure but maybe this is what you're looking for? internet-map.net/

I seem to remember from my networking class that the prefix's are sorted according to length in descending order such that the first match is the correct match.

You'll notice the prefixes also contain a slash, followed by a number. That number is the key. It tells you how many bits long is the prefix. For IPv4, where the addresses are 32 bits, what happens is: 1. Start off with the assumption that the matching prefix is 32 bits long. 2. Are there any prefixes this long in the table? 3. If so, get the one that matches it. 4. If not, assume the prefix is one bit less, and go back to step 2, unless you've already checked 0 bits, and there's no match. 5. If there's no match even in 0 bits, drop the packet (since it's destination is unknown). Now... Many modern routers also go one level deeper, in that they allow you to have the same length prefix defined multiple times, once for each interface, thus allowing you to reduce the overall load on the network - you send one packet on one interface, and the next on another, thus allowing the two other routers to process more packets (i.e. not just yours). EDIT: It's important to note that for the sake of optimization, many routers "sort" the table in advance, upon creation/modification, and merely check records in it in a single complete pass.

boenrobot Tyranisaur As long as routers _behave_ in the way described in the video they can interoperate, so they're free to _implement_ that behavior any way they choose. I'm sure different routers do it slightly differently. The thing that jumps immediately to mind is building a bitwise trie (prefix tree). When you reach a terminal state in a search of the trie, that nodes is necessarily the longest prefix for the search parameter (IP address, in this case). This would make runtime independent of the number of addresses stored in the trie. However, it's possible they have something tricky they can do with hashing that provides an even faster implementation. It's also worth bearing in mind that routers aren't (normally) general purpose PCs, and the big routers in particular are highly specialized machines. That means that they may include hardware specifically designed to accelerate the operations needed to do forwarding table lookups.

It depends on what role you play on the internet. If you had two ISPs connected to your home router you could simply insert routing table entries manually pointing a specific route at one of the two interfaces connected to those ISPs. If you're an ISP you can decrease or increase BGP local preference, or manipulate other metrics BGP uses for best path calculation, to manipulate which route is chosen as best. The key takeaway here is that you can only manipulate the routing decisions that happen within your own network. You can never arbitrarily choose what path a packet should take, only what the next "hop" for that packet will be.

It and the routers it would talk to have to be configured. OSPF will not communicate with an arbitrary router; in fact there is a lot of configuration that needs to be identical for neighbour relationships to form (even Hello timers), not to mention encrypted authentication. Why? Because otherwise, you could simply connect a rogue router with a bunch of extremely-specific addresses that point to nowhere and flood a network with garbage routes in no time. Worse, the infiltrator would have ALL the information about ALL the networks and endpoints known by the corresponding OSPF process (in short, the routing table)!

When you take a new router out of the box and plug it in, it immediately does absolutely nothing at all. You have to tell it about your network. That is, what interfaces you want to use, what networks those interfaces are responsible for, and how it should advertise what it knows to the other routers. If you plug in your "router" you bought from your local consumer store, it politely asks your isp where it should send all of the packets not relevant to it's subnet. I have no idea how they did a video on routing without even mentioning gateway of last resort. What they failed to mention about the routing table is that routers are usually configured with "and if you reach the end of your routing list, send it here, it's not your problem anymore." That's known as the gateway of last resort, or simply, gateway. Every network device on network maintains a routing table, if you want to see the one for your computer running windows, open a command window and type route print -4. What your computer does when you type a web address into your browser is check that ip against it's list, running down until it finds a match, then, when it hits the last one it goes "aha! here's where I send this" and off it goes. Then your home router takes a look at it and checks it against it's routing table, finds it doesn't have any idea where it goes, and sends it up to your ISP. That's where the magic of routing protocols really happen.

^ A couple of errors: 1. A router doesn't have to have a default route, it can simply discard packets it doesn't have a matching route for. In fact this is the recommended practice; using default routes or default gateways are only common in consumer environments. In fact, this is the primary difference between a router and a switch: while the latter is designed to broadcast a frame for an unknown destination, the former is designed to drop packets that aren't a match. 2. "Every device on a network maintains a routing table" - not true. Only Layer 3 devices that participate in routing & forwarding do. For example, (L2) switches do not contain a routing table (most switches these days are L3 capable, for line-rate routing between VLANs and other purposes). 3. The computer doesn't have an exhaustive list of all IPs on the internet - that's equivalent to running full BGP on a desktop machine! Theoretically it has a hosts file, but this isn't really used in practice any more; the OS just sends a DNS request out: the router forwards it without caring. Once the DNS response is back, the OS can then initiate a connection with a specific IP address: the router again forwards it without regard for what it is for. From a consumer-grade router's perspective, it doesn't know which end is the "ISP", it's just got two interfaces with two IP addresses, possibly one public. It simply maintains its own routing table between these two.

Karn Kaul I'm trying to understand the relation between routing tables and the link state protocol. Where the actual search in the routing tables comes in? It feels like two different ways of solving the problem of "where to send". When the router receives a packet, it'll then get the match with the highest subnet mask or it'll apply the link-state protocol? I might be a little confused here, so i'd appreciate your answer!

Karn Kaul Network device. Please tell me what layer 3 is called in the OSI Model. Whats that? It's called the Network Layer? Well then, guess it wasn't an error so much as semantics. I'll say it again, this time being perfectly clear. Every Layer 3 device on the network maintains a routing table. Full Stop. If you don't believe me, check it yourself. In the Windows Command Line: route print -4 Hosts file doesn't even play into this. That's DNS, you are jumping around the model. DNS is at the application layer. Remember, if you can interact with it with human language, it's an application. From a consumer grade router's perspective, it knows exactly which end is the ISP. It's the end with the default gateway. GOLR is the route it uses for networks not under it's responsibility. Grossly oversimplifying it, the side it NAT/PATs over is the ISP. If we are talking pure routing, in the perfect beautiful world that doesn't exist, you would be correct in saying it doesn't know which side is the ISP. We don't live in that world.

nothing so dynamic since the ARPANET days-- the way i illustrated it made it look like hop counts (ie, every link "costs" the same amount). in practice, links are actually assigned integer weights when the routers are configured, and those give the costs. (the weight for a link may well be initially set based on some combination of the delay and capacity of the link though, and then subsequently tweaked to get the traffic flowing in your network how you want.)

No, not really. Proxying is another topic unto itself, and forced proxying even more advanced than that.

ISP's etc can cache KZfaq videos. It's part of KZfaq's system to reduce strain on the internet and their network. But ISP's being ISP's they can mess it up and are somewhat prone to failures. There are ways to disable it and avoid your ISP without a VPN. Look up some browser extensions.

Dykam Alright, thanks! It's not something that only happens with KZfaq, so that will definitely be of help.

Richard Smith I'm pretty sure it's finally been ruled illegal, here in the states, to force companies to pay more for bandwidth. And it does irritate me when that happens, thus the VPN (one of many reasons I got it). Of course, either way I know that it's not KZfaq now, so thanks for your reply!

ssholum The caching isn't illegal (it is in cooperation with KZfaq), and AFAIK the practice of capping I think still is legal.

Dykam Not the caching, the throttling. Pretty sure a court ruled that Comcast's requirement for Netflix to pay for equal priority on pipes was unlawful, setting the legal precedent. Of course, I don't use Comcast, but it shouldn't take much more for it to be considered unlawful to extort companies in such a way (assuming this common carrier thing ever gets concluded).

FEARbraveheart What do you find baffling about X11?

rooter* to give someone a root... shag... sex.

Rooting is sex. It just sounds the same.

www.google.com/

In Australian slang to "root" means to have sex (similar to the word "f***").

oh, of course

(row´ter) (n.) A router is a device that forwards data packets along networks. A router is connected to at least two networks, commonly two LANs or WANs or a LAN and its ISP's network. Routers are located at gateways, the places where two or more networks connect. Routers use headers and forwarding tables to determine the best path for forwarding the packets, and they use protocols such as ICMP to communicate with each other and configure the best route between any two hosts.

Someone who has sex, or obtains privileges on a a unix or linux system. This video however is about routers.

Might be throttled by your ISP.

Comcast wouldn't care to throttle that. i think its KZfaq. plus why would my ISP throttle KZfaq and not other sites?

bulldogg110 Because KZfaq competes with Comcast's own cable services. It's very unlikely it's KZfaq, KZfaq has no idea whether you are using a VPN or not. You probably connect to the VPN, which is fine for Comcast, and then the VPN connects to KZfaq, on their own non-comast ISP.

I don't connect to a virtual private network. switching DNS servers wouldn't work either. meh, its probably me, but KZfaq is known to go under maintenance often.

bulldogg110 Ah sorry, I had a conversation about this earlier last week and I guessed I mixed up the conversations. I never have any issues with KZfaq, and I'm on here *a lot*.

192.168.127.0/17 is 192.168.0-127.0-255, i.e., the same as 192.168.0.0/17. (127d = 0111 1111 b) The reason the latter prefix notation is "better" is that, by convention, any bits that are not part of the prefix are made 0. However, they both indicate the same range of IP addresses.

Why, the 17th bit is the first bit of 127 = 0 and 128 = 1, and we only look at the 17th bit. How could one bit have three different states?

Right. The 17th bit of 192.168.127.0/17 is 0, and the 17th bit of 192.168.0.0/17 is also 0. Therefore, they are the same prefix.

Yes, if you're talking about strict CIDR subnet notation, 192.168.127.0/17 does technically indicate the lower half of the subnet, but it's sloppy notation because it's a random address in the middle of the subnet and not the actual network address, which is 192.168.0.0/17. The way he wrote it is no different than 192.168.43.177/17. However, many devices don't automatically calculate the network address, and they will throw an error if you try to plug in 192.168.127.0/17. Some will also assume a range by calculating how many hosts can be in the given subnet mask and add that to your input, therefore 192.168.127-254.

Thought that's what I said. :)

This is still the case with IPv6.

That's the way it's supposed to be. Though I fear people used to the natural firewall effect NAT has and forget to set up an appropriate firewall. IPv6 has unroutable addresses as well, so that's not exactly the case.

Mostlyharmless1985 While that is true, in general you can. There have always been exceptions.

Very

context?

The same way Americans can differentiate between rout and route - context. In fact, rooter is also a common American pronunciation as well.

Yes, but IANA has several "private" blocks, some of which are supposed to be used for for documentation (and by extension, examples like this one). tools.ietf.org/html/rfc5737 So with a little studying, the presenter could have used those instead. A lot of people know the 10/8, 172.16/12, and 192.168/16 prefixes, whilst the other ones like 192.0.2.0/24 are not talked about as often.

amigojapan It was Pakistan trying to block KZfaq in Pakistan.

Kasper Guldmann this seems like a terrible vulnerability that the internet may have against an attack from another country.... in virtual warfare....

So which one is it?

Frostbite1003 The one over yonder, god damn it.

en.wikipedia.org/wiki/Subnetwork

Thanks for the cite.

+Jim Fortune Yes, I think they really should do a subnetting video. It's what TCP/IP relies on and people may well have come across it as part of their IP address on their computer. It's certainly been seen by the average user far more than BGP, RIP or OSPF will have been. It's neat how it works.

...except noone but Google has KZfaq's private TLS key. Therefore browsers all over the world would warn their users, hey, you're not actually connecting to KZfaq like you thought you were. The digital signature isn't right. But they could execute a wicked DoS as the video and you propose. The security in the Border Gateway Protocol (BGP) is fairly lax and depends on peers trusting each other. There are also filter mechanisms to prevent propagation of false routes. If for example I've been delegated two netblocks, say 24.48.72/24 and 24.48.96/24, and I advertise I can handle traffic for 128.205/16 (SUNY at Buffalo), my peers (probably one or two ISPs) can disallow propagation of that, or any prefixes other than the ones they know can be reached through me. I might do that to set up load balancing, overflow traffic, or redundancy in case of link failure. In practice though, no IPv4 prefixes longer than /19 are propagated via BGP, according to ISP policy...so those example network prefixes above are too long (for BGP).

That's a very good point about the TLS key. Thanks for the info!