Map of the Internet

Wellington Grey writes "Author of the popular webcomic xkcd has put up a hand made map of the internet as today's comic. He also has an interesting blog entry detailing some of the work that went into it, such a pinging servers and creating a method of fractal mapping to display related regions as contiguous sections on the grid." The drawing is pretty damn impressive; somebody get on making that thing a giant wall poster so I can paper over Taco's office door.

Comments

Rasterizer.

[celardore]

The drawing is pretty damn impressive; somebody got on making that thing a giant wall poster so I can paper over Taco's office door.

Have you tried something like Rasterizer?

Re:Rasterizer.

[Council]

To everyone who's asked for a large poster of this -- I'm going to be offering large prints of it in the xkcd store before too long, but for a handful of reasons I can't easily do it immediately (I'm in the middle of the holiday rush with shipping out t-shirts). It's cool to hear so many people are interested, though! Thank you!

I would actually like to see someone else create a computer-generated poster with a higher level of detail (there will be algorithms for the mapping on the blag soon). I think you can do some interesting things with this fractal; it'd be neat to see all the websites you visit marked with red dots, more detailed survey info for the registry patchwork, server density/space usage (the 63-74 blocks are more densely populated than anything else), etc.

Re:Rasterizer.

[ei4anb]

obligatory reference to the CAIDA maps: http://www.caida.org/analysis/topology/as_core_net work/

I realy do like the simple structure of the xkcd map though; like the London Underground map it is a simple representation that took much work to make it so simple!

xkcd

[Tet]

xkcd is a work of genius. See, for example, this classic.

Re:xkcd

[Fordiman]

That was special. Almost makes me want to install Linux on my girlfriend, just so I can try that.

Meanwhile, I agree with the killing Ann Coulter thing. She just makes humans look bad.

Re:xkcd

[Thornae]

Note that there is no request for a password.

The implications of this are left as an exercise for the reader...

Be warned: If you're viewing xkcd for the first time, you might end up reading through all of them. It's simple but brilliant.

Re:xkcd

[erpbridge]

Methinks the Girlfriend is insecure? Seems she is easy to root.

Since the girlfriend takes commands over the air, that makes her an open access point?

Re:xkcd

[loconet]

Here is another hilarious one.

Re:xkcd

[finkployd]

Yet none are as funny as this

Finkployd

Re:xkcd

[Phleg]

I find it interesting that most of the people who see the comic immediately assume a boyfriend/girlfriend relationship between the two. If you look carefully, nowhere does the comic indicate which gender either character is.

Clever

[inviolet]

Wow, I wish I was clever enough to come up with stuff like this.

The author gets additional Cleverness Points for thinking to post the geonetric locations of the major geek sites (slashdot, digg, boingboing, etc.) in order to encourage those sites to repost links to the author's website.

Re:Clever

[strider44]

It's not like he gets ad revenue. XKCD is the best computer comic I read and I don't really think he craves the attention so much.

Real Map of Internet

[Delta-9]

Thats neat, however opte.org is working on realtime maps of the internet.

Re:Real Map of Internet

[grommit]

That's nice, however those opte maps don't show the same information as the xkcd map does. While a whole bunch of lines randomly spread around has a certain spartan appeal, it doesn't convey any information. I can't look at the opte maps and say, "Oh, there's so and so" or "here I am." So, I'd hardly call them maps. Maps usually have information tags describing/naming places. Maybe those LGL files contain that information? It'd be nice if they made screenshots of the output of those LGL files though.

Interesting...

[$RANDOMLUSER]

But where's the "Here there be dragons" part?

Re:Interesting...

[Savage-Rabbit]

But where's the "Here there be dragons" part?

In view of the way humanity's moral compass has been recalibrated since the middle ages I think the need for the creation of a "Here be porn!" annotation is more urgent.

Re:Interesting...

[ei4anb]

The Dragons are shown in real time on this map http://isc.sans.org/large_map.php

Amazing web commics

[TheRagingTowel]

What amazes me most is his ability to make you see the character's face expression although it's a faceless stick figure (eg this). That and that he seems to be an absolute geek :)

MIT

[minus_273]

I always laugh at how MIT half as much as all of latin america and as much as all of Africa.

I remember being in MIT and getting a real fixed IP for every single device. We actually had a coke vending machine that was hacked and online with its own IP. Considering they has so much that they are no where near running out, I'm sure there are a ton of toasters online at MIT as well.

Re:MIT

[Anonymous Coward]

I always laugh at how MIT half as much as all of latin america and as much as all of Africa.

Buh?

We actually had a coke vending machine that was hacked and online with its own IP. Considering they has so much that they are no where near running out, I'm sure there are a ton of toasters online at MIT as well.

Wuh?

Re:MIT

[minus_273]

I haven't had my coffee...

Re:MIT

[Pasquina]

Each dorm is assigned all of a second-level IP: 18.XXX.*.*, that's 65536 IP addresses per dorm. At about 300 students per dorm, that's more than 200 static IPs per student...just in case. My fraternity is assigned 512 IPs for 45 guys.
If nothing else, it has skewed my opinion on how quickly we're running out of IPv4 addresses.

I've also heard that MIT rents some of their IPs to Portugal. (This was also the subject of a supposed hack that some MIT student took out an entire country's internet service for a little while.) Does anyone know if either half of this is true?

Too much time

[OhHellWithIt]

Someone obviously has too much time on his hands. And to think he could have been reading /.

Risk?

[onetwofour]

Anyone fancy a game?
Good news is that we could wipe out the USA quite quickly.

Good job, but...

[level_headed_midwest]

They did a good job in labeling things like local, multicast, loopback, and VPN addresses, but they forgot to note 169 as such.

Re:Good job, but...

[dpilot]

Since you're mentioning the zeroconf space, it's also worth noting that they pointed RFC1918 at "192.", though I guess they discounted everything but "192.168.", but in the meantime they completely forgot about 172.16.-172.31. and gave 10. to cable companies.

I've been thinking for some time that 172.16-31 might be a better place to hide my LAN, away from normal expectations. In a very meager way, this confirms it.

Re:Running out?

[revlayle]

THAT what must be clogging the tubes. Not porn... GRASS!

IPv4 space

[JohnnyBigodes]

I thought we were (supposedly) running out of IPv4 space... but the map shows quite a few unallocated blocks. What gives?

Re:IPv4 space

[forkazoo]

I thought we were (supposedly) running out of IPv4 space... but the map shows quite a few unallocated blocks. What gives?

Look at how much spqace MIT has. Now, look at how much space the whole of Africa has. Even if we assigned every last block, we would probably never see an African university with a whole /8 to itself. Think about how many people are in India and China, and compare the asian assignment vs. the US assignment. It will be impossible to ever make IPv4 fair. IPv6 allows us to just bypass the whole issue and let everybody have as much address space as they could possibly use.

Re:IPv4 space

[wayne]

Actually, wikipedia has a very good summary of when IPv4 address space exhaustion will likely happen. In particular, while the IPv4 allocation graphs made by Geoff Huston aren't as pretty, they are likely far more accurate than xkcd's. The only problem with Geoff's predictions is the exhaution date keeps getting moved forward so his dates are probably best-case predictions.

Basically, yes, the IPv4 space is running out. It is still 3-5 years out for IANA exhaustion and further for the RIRs and ISPs, but it is something that people need to start planning for. The predictions about IPv4 addresses running out back in the 90s was before the development of things like CIDR allocations, NAT, RFC1918 private network numbers, HTTP1.1's virtual hosts, DHCP, and the dot-com crash. There haven't been any new "gee, we can make the IPv4 space go a lot further if..." type ideas for years and it doesn't appear likely that any more large savings will happen before it is too late to deploy them.

Re:IPv4 space

[forkazoo]

This sounds suspiciously like "640K ought to be enough for anybody."

Have you looked at how many IP's you get in IPv6? Seriously, I once saw the number and it took me several minutes of googling to figure out how to say the number outloud because I had never encountered a number that large. Given that IP will only be useful for a single planet network, we should be good for a very long time.

Quickly googling, I saw these explanations of how many addresses we get with IPv6:

(667 sextillion) addresses per square meter

3.4 times 10**38 addresses, or 5 times 10**28 (50 octillion) for each of the roughly 6.5 billion people alive today

I'm perfectly comfortable being quoted saying that 50 octillion addresses ought to be enough for anybody. (Considering the whole of the current IPv4 Internet is only 4 billion some odd addresses...)

Dragons?

[Marbleless]

How boring our world has become.

Old maps used to claim "Here be dragons", but today it is "Unallocated blocks".

Where has the mystery gone? ;)

So why

[dattaway]

does a company like Halliburton get a whole square? Are they planning to invade others?

Re:Where's the money?

[masklinn]

Or is there pure geek value in this?

I take it you've never read xkcd have you?

Use Domains+Web Sites, instead of IPs?

[Kadin2048]

Although a map of the IP address space is probably more interesting and informative, something that was based on the distribution of domain names might be more appealing to a non-technical audience; perhaps something showing the relative size of various sites beneath each TLD, with some factor based on popularity and grouped by semantic distance and interlinking.

E.g., so you'd end up with something that had big regions for the major TLDs, and then within them you'd have semantically related regions (sites that are related based on keywords or link to each other heavily). The base unit could be sites, and their size would be proportional to their number of publicly-accessible pages times a 'popularity factor.' Maybe you could extract some of the popularity information from Google (not that they'd probably like you hitting them with a lot of scripted searches).

I think it would be neat, particularly if you ended up with something that showed such locales as the Spamblog Ghetto, Fortress Corporate America, and, of course, the Porn District.

Hilbert curve

[fbonnet]

FYI he uses a Hilbert curve to map the IPv4 space on a square. This is simply brilliant, elegant and beautiful, clearly the best map of the Net I've seen in years. I love how the range of Multicast IPs renders as a square.

http://en.wikipedia.org/wiki/Hilbert_curve

One Factor

[Kadin2048]

In reality, the security of the girlfriend system is hardware-based; it requires the presence of a specialized dongle.

DEC?? I think not

[Necron69]

I have news for this guy. DEC (net 15) hasn't existed in nearly a decade, and HP and Compaq merged like four years ago. So Nets 15 & 16 should be labeled "HP".

All your IP space belong to us!!! Bwahahahaaaaaa!!!

- Necron69

A good reason to move to IPv6

[Rocketship Underpant]

Isn't it kind of sad that the entire continent of Africa gets the same number of IP addresses that Prudential, an insurance company gets?

Re:A good reason to move to IPv6

[iabervon]

Latecomers to the internet, like Harvard and Africa, have their networks structured such that they don't need huge numbers of IP addresses. When MIT originally set up their network, their routing was done by IP address block, so the routers could all decide where to send packets based on a single octet. So, if you have one computer in a location without any other computers, it gets 65536 addresses. Furthermore, the original routing between sites was simplified greatly by having the first octet dictate which site would get the traffic, so it would have been very difficult to give MIT or Federated less than 16777216 addresses, because traffic for all of those addresses would be routed through a single link to the rest of the internet.

These days, any infrastructure device is perfectly capable of looking up addresses in a table, and can discover and store the mappings for your whole network with no trouble at all. With this sort of hardware (which is all that's still available), each computer only needs one address. When there are 16 million computers in Africa on the internet, they can have more addresses. For that matter, MIT would give up most of their addresses if there was a shortage; last time I checked, only 18.*.0.* was generally used.

The real reason to go to IPv6 is not that there aren't enough addresses for everybody, but rather that there aren't enough addresses to not have to worry about allocation. With IPv6, every NIC that'll ever be created can have its own IP address (based on its MAC), plus addresses it gets by being connected through a router, private addresses, loopback, and so forth. There are useful effects of having so many addresses total that you can assign large spaces of them for purposes other than just having an address for each device on the internet.

IPv6 is there too...

[scsirob]

Just float your mouse over the picture and he will tell you what the IPv6 version looks like.

Even more clever, and sooooo right ;-)

Tubes?

[j00r0m4nc3r]

I don't see any of the tubes

Re:Why was 192 picked as private?

[rednuhter]

in binary 192 is 11000000
so with bit masking it makes sense.
http://en.wikipedia.org/wiki/Bit_mask

Re:Why was 192 picked as private?

[Andy Dodd]

192 = 128*1.5 or 128 + 64

i.e. while not strictly a power of two, it is closely related to one.

More specifically, the bit pattern for 192 is a nice clean 11000000

oblig.

[Bugs42]

Almost makes me want to install Linux on my girlfriend, just so I can try that. You know, of course, what this means.

Slashdotter1: Dude, I met the most awesome girl last night! She's hot, funny, smart, AND a gamer!

Slashdotter2: Yeah, but can she run Linux?

Re:oblig.

[Rheingold]

Now imagine a whole clone army of Natalie Portmans running Linux, serving up hot grits.

Wow, that was so 2000.

Re:oblig.

[dosquatch]

Now imagine a whole clone army of Natalie Portmans running Linux,

So, you're suggesting we imagine a babeowulf cluster of these?

Useful

[Hegh]

That's actually quite useful to me. Twice I've watched somebody attempt to brute-force their way into an FTP server that I run for myself (which I have since taken off of the public internet, since I realized I only use it on my LAN), and now I know that the attacks which came from 61/8 and 62/8 are in Asia and Europe, respectively (therefore I don't have to worry about blocking those entire IP ranges, since if my FTP server were public again, I would never be in one of those ranges trying to get in). Anybody else have a practical use for this?

Re:private ranges all marked differently?

[AFCArchvile]

The private, nonroutable IP ranges, according to RFC 1918 are:
10.0.0.0 to 10.255.255.255 (10/8 prefix)
172.16.0.0 to 172.31.255.255 (172.16/12 prefix)
192.168.0.0 to 192.168.255.255 (192.168/16 prefix)

Internet map from Wikipedia

[Inyu]

THE MAP: http://en.wikipedia.org/wiki/Image:Internet_map_10 24.jpg

AUTHOR'S NOTE:

I created this small partial map of the Internet from the 2005-01-15 data found here using a slightly different rendering technique than was used to generate the maps there. Each line is drawn between two nodes, representing two IP addresses. The length of the lines are indicative of the delay between those two nodes. This graph represents less than 30% of the Class C networks reachable by the data collection program in early 2005. Lines are color-coded according to their corresponding RFC 1918 allocation as follows:

• Dark blue: net, ca, us
• Green: com, org
• Red: mil, gov, edu
• Yellow: jp, cn, tw, au de
• Magenta: uk, it, pl, fr
• Blue-green: br, kr, nl
• White: unknown

Big BIG HUGE (probably unusable in articles) version can be found at Image:Internet map 4096.png.

Re:Beeb

[Andrew Kismet]

http is the killer app of DARPA's platform.
The British deserve a pretty damn sizable chunk of it, with respect to population and usage.

DNS?

[Kadin2048]

is there a simple way to get mappings from domains to IP address space--in bulk?

Erm, I don't know of a publicly-available list, but it seems like it would be pretty easy to generate one by just using DNS queries.

What you're asking for is pretty much the function of the DNS system, after all. You could easily write a script that took a list of domain names and resolved them to IP addresses -- you'd just want to make sure that your upstream DNS provider didn't block you for being abusive or for looking too much like a DDoS.