Slashdot Mirror
Mapping the Internet
triple6 writes "No, this isn't spam. Bill Cheswick of Bell Labs has been
mapping the entire Internet and then plotting the results, in color, on paper. The images are, well, stunning. Now, you can buy a poster of one of these images from Telegeography. The poster might be a bit pricey to use to decorate your dorm room. Personally, I'm going to wait for the t-shirt. " I just wish I had a little arrow that said "You Are Here". Those things are super cool.
I stand corrected then. While trying to get my head around your explanation, I did a little digging and found an excellent primer on HUD.
:)
Is there a formalism to explain the effect of observation on the observed? After all, it's a serious issue even in such non-technical areas as psychology. I would think that someone would have branded it with a name. I vote we call my (mis)definition the Jabber Uncertainty Principle.
Hmm... Maybe I can mis-quote Newton, and apply action-reaction to sociology, in an effort to explain the partizan alignment dichodomy between the Executive and Legislative branches of the U.S. government. Or maybe to explain the U.S. and French revolutions as reactions to oppressive and tyrannical rulers.
-- What you do today will cost you a day of your life.
Where'd I leave my old physics book? Heisenberg may not approve, since (I agree) his work was on the quantum level, but I think my invoking HUD still holds because in mapping the net we're not dealing with macroscopic physics.
My point in bringing HUD into it was that a comprehensive mapping effort would in effect be a denial-of-service, with lesser undertakings resulting in more accurate, but less precise models. It's akin to finding the bowling pins using a bowling ball. In measuring the network traffic, we displace that traffic. And in mapping the routes, there is a probability of changing them. How is this different from looking for electrons with a beam of light?
Well, we're not dealing with subatomic particles which have a wavelike location probability; but the thrust of it seems the same. We are dealing with packets, which have some probability of of taking some route between two endpoints. If we're bombarding the possible paths with our own packets (traceroute), we're bumping the observed ones onto routes they may be less likely to take if we were not making any observations.
Looks like a decent topic for a research paper, no?
-- What you do today will cost you a day of your life.
I can see my house from here!
Another damned comic
+++ NO CARRIER
Site's dead. /.ed no doubt.
/. actually cached/proxied/copied these slashdotable sites rather than just linking to them? Hmm?
/., everyone gets to read the page and there isn't a trail of dead servers left in the /. wake.
What if
The traffic stays within
Deleted
I went to a talk on this subject by Bill Cheswick this past winter. The pictures that he showed were cool. He showed the different levals as they were coming out of his computer (first hop second hop and so on) He did give away some of these maps but i did not get one.
Um, I can't seem to access the first link in the story (guess it's been /.:ed), but perhaps An Atlas of Cyberspace (cheezyness alert!) is something similar. Pretty cool images there, too.
main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
CyberGeography mirrored: here and here
-Yarn - Rio Karma: Excellent
Since the site is slashdotted, if he tries doing an updated map now, its going to be a pretty small map.
Bob.
Very funny. How about stating it this way...
The action of observing traffic patterns on the net changes said patterns, thereby rendering the findings inaccurate. There is an inverse relationship between the attempted precision of the observation and it's detrimental effect on the results.
In real simple terms, you can only measure network delays by adding to the traffic, thereby increasing the delay. You can only traceroute by adding to the traffic, thereby potentially causing a load-balance change in the routes.
The act of observation changes that which is being observed. Sounds a lot like the Heisenberg principle, applied more broadly, doesn't it?
-- What you do today will cost you a day of your life.
There's probably a legal line or two that would make duplicating (even mirroring) illegal... especially when you get into the fact that many of these sites generate revenue from banner ads. Are you going to duplicate the banner ad functionality too in your mirror?
It gets complex, and he certainly couldn't do it without their permission...
Though it's a great idea. Maybe we could all just use a common proxy.slashdot.org that proxies these things. I don't see a problem with that...
the search engine at www.google.com does that for most of its pages, and it seems to have no legal issues. They have a little "cached" thingy next to most of their links. The reason for this is that any search engine database is usually pretty out of date, and if the page no longer exists you can at least see from the Google cache what it once looked like.
/.ing itself. A couple of icons and a massive amount of text thing, but if they also had to handle shipping out copies of these huge map images of the internet to ungodly amounts of people while the real map image website was down.. well, could it handle it? Maybe the caching would be text-only, and have the BASE HREF set accordingly. Then you could at least see an imageless version if the original site gets slashdotted, and it would solve the banner ad problem as well.
your point about banner ads is a important though--maybe the caching mechanism could be written in such a way as to recognize banner ads (the way iCab and some filtering software does..) and to make those images direct links to the orignal file instead of cached?
a better question would be how Slashdot itself would handle it without
anyway the point is, google.com kicks ass.
Irritable, left-wing and possibly humorous bumper stickers and t-shirts
how difficult would it be for these people to distribute their maps in a vector-based image format instead of bitmap? would it be possible? it's all colored lines anyway. It might be large because it's over 100,000 colored lines, but you'd at least be able to zoom in. And the bitmap method makes it hard to see individual sections of the map..
what format would work for this? are there any formats that would allow you to name individual points on the map, so they could distribute along with the map a [probably huge] automatically generated index of the IP adreses?
would that kick ass or what?
Irritable, left-wing and possibly humorous bumper stickers and t-shirts
Wonderful pictures. But `coloured by IP address` isn`t enough. What colour represents what address? It would be cool to know, cause that way you`ve got a better idea of which part of the map you`re actually in..
WOW. that is really cool looking. i wonder how he handled multihomed hosts and isps.
:) hats off to the men who do things!
he more than likely didn't though. with bgp you will end up seeing a particular host on the internet only through one of their particular links.
the only really true (an impossible) way to map the internet is traces from all hosts to all other hosts. that way you get a full view of the picture, not one that is distorted by being on one node of the internet.
because of this one problem i think his maps should look like a tree and not a star. with a tree you get the more realistic view of only ONE viewing point.
that is still better than any internet maps i have seen before
signatures are for fools with hands
Ding, dong, the Netscape server's dead!
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
I have to ask.. How does one capture the state of the dynamically routed internet in a meaningful way?
I was the blurb about generating the images over time, and viewing them as a film, with each image being a frame - but that would make for little more than a documentary of what was. Interesting to see the patterns and all, but not really a reference.
Further, doesn't the Heisenberg Principle make comprehensive mapping of the net a bit of a paradox?
-- What you do today will cost you a day of your life.
I bet they're (peacock maps) feeling pretty stupid that they designed their web site almost entirely using *images* of text instead of simply plain old text.
there are other issues too. I've looked at this map and others at cybergeography before, and none of them seem particularly useful.
One of the primary problems (other than the one you mention) is that on cheswick's map, all "roads" are equal. there's no differentiation between the OC-48 "freeways" and the T-1 "dirt roads". (hmmm, can I do anything else here to perpetuate that assinine highway metaphor?)
so, all the map really shows is lots of connected points in different colors. even though it shows a valid representation of real data, it is not effective in conveying information about the net (unless you're trying to figure out which ISP has the most nodes).
This is another one of those first steps. How long before they say "damn. screw this 2d mapping stuff, we need more space, 3d time" and we have a visual representation of the net. Then we start seeing web browsers come out with 3d web surfing (need a new word here people) and then someone spots the power of 3d goggles. Before long we get the cool, jack my head in and the data starts to flow over like ICE. Oh yer.. pure data, hard data. Jack it in, let it flow out.
How we know is more important than what we know.