According to an article in the San Francisco Bay Guardian, the design problems were known 40 years ago. As you say, it wasn't "fine when it was built" - it's not like tigers have gotten bigger in the last century, though perhaps Siberian tigers are bigger than whatever species of tiger they originally put in it. And the moat was never deep or wide enough, and tigers are even better at leaping across than jumping straight up.
The real reason the wall worked that long is that none of the tigers had previously felt motivated enough to jump at it. Apparently Siberians are more aggressive than Bengals, and maybe the two drunk kids pissed her off or just acted enough like prey or cat toys that she went for them. My cats sit on the couch looking out the window at Bird TV, and when one of them sees a laser pointer red dot he has to jump for it without thinking about it first (the other one says "Hey, stop wavin' that thing around".) And I've seen zoo leopards looking at the crowds, intensely tracking the smaller ones that get separated a bit from their herds; I'd feel much safer around tigers.
The number of tigers in the wild has been declining rapidly, and if you want real numbers check reputable sources. Last I heard it was under 3000, and it might be a lot fewer by now, and general estimates are that by 20 year from now they'll be extinct in the wild.
The number of tigers in zoos is about 4000.
As many as 3000 tigers may be in farms in China, being raised to sell as traditional medicine for people whose penises aren't big enough or who think their bones will make them stronger.
The number of tigers that are kept as pets by Americans is about 6000. There are animal activists like Tippi Hedren trying to make laws against keeping tigers as pets, because almost nobody who has pet tigers has enough space and resources to let them live like tigers need to, especially the occasional drug dealer in some apartment building in New York who wanted to out-macho his competitors' pit bulls. She's well-intentioned, but the species needs all the genetic diversity it can get, even though tigers aren't meant to live like house-cats.
There was a tiger attack at some animal park a decade or two ago, and some TV reporter asked the trainer whether they'd known the individual animal was dangerous before the attack. His reply was "Maam, they're tigers."
Or as a friend of mine commented, "If they were six-foot cuddly bunny-rabbits, we'd have called them bunny-rabbits, not tigers!"
There were actually two cables cut - FLAG and SMW-4, and according to one article they were cut in different places.
Also, if you look at how internet transmission works, while you obviously want geographical redundancy, that doesn't mean that you don't send traffic on all available routes. Carriers are going to make sure they've got enough redundancy for their critical load levels (e.g. the voice network and private-line customers), but if they're doing redundancy at Layer 3 they're going to send traffic across multiple routes because it doesn't make sense to leave them idle.
To some extent, if you're doing Dense Wavelength Division Multiplexing, and if you haven't lit up all your wavelengths (because the optics and routers at the end are expensive), you can sometimes divert some wavelengths to the alternate routes. For instance, you'd provision wavelengths A, B, and C on the west route and Z, Y, X on the east route, and if something breaks you can push them onto the other route. But once your cable fills up, you've got less ability to do that until you build more cables or put even more expensive optics on the ends to light them.
And sometimes you just get surprised - like the big Taiwan earthquake last year that took out N-1 of the undersea cables between northern and southern Asia, which almost all go between Taiwan and the Philippines since that's what the ocean floor shape makes you do. They were spread out far enough to avoid problems with ship-anchors, but the quake was over a wide area. And there was a quake in the Med a couple of years ago that took out more than two cables as well.
Pretty much everybody in the telecom business enjoyed reading it.
Re: your last point, remember that the WTC was attacked twice, the first time with a truck-bomb in 1993, which wasn't successful in taking down the building (because some fscking rude New Yorker had taken the parking place next to the pillar they'd planned to park the truck in:-), but which caused a lot of damage and panic. A friend of mine's father was working in one of the top floors when in happened, and messed up his back enough during the evacuation that he decided to retire early; otherwise he'd probably have been working there in 2001.
Yes, they're registering thousands of names that way - most of the names in DNS are getting tasted once or kited repeatedly; I've seen estimates up to 90% of transactions..
The reason that most of them do it is ad-banner revenue - if enough people hit your web site by typing keywords into their browser that you can pay for it with ad banners or Google adwords or whatever, then you're making a profit. Domain tasting lets you try out a name for nearly-free and see if it generates enough hits over a couple of days to make it worth keeping. Depending on how much revenue you get, it *might* be worth paying the whole $6/year to keep it, or it might only be worth the cost of kiting it, which is still non-zero (it's your cost of money on $6/year), and which risks somebody else grabbing it if their automated tools are faster than yours.
Grabbing names that somebody else is checking the ownership on does two things - one is that it's sometimes possible to make money by scalping the name to them, but also it means that the name had *some* probability of being something that humans might type into a browser, so it's more likely to be worth grabbing for the advertising revenue. It's especially common if a domain name has just expired, because there's a good chance that the site had some traffic on it that you can capture for your own ad banners. A friend of mine just had her domain name stolen after owning it for 10 years - hopefully she can get it back but at least as of last week it was pointing to a domain-parker page.
There's a whole parasite-infested industry called "domainers", who are people who buy up domain names for the advertising revenue and for the possibility of making a profit by selling them either to other domainers or to businesses that actually want the name. They're parasites because they're using name space without providing any useful content, and in many cases they're also filling up Google with pages pointing to their pages, so that Google searches are more likely to get their sites than the interesting stuff you were looking for.
The practices are really pretty much the same thing; if one parasite stops kiting a given name, another parasite grabs it, so it doesn't really matter.
The original domain name grace period rule was ostensibly to prevent people from complaining that they'd paid $35 for a domain name that they'd mistyped or that somebody else claimed was their trademark, and it avoided trademark disputes because you could simply return the name rather than having to sell it to the alleged trademark owner who might then complain that you were trying to name-squat them. It's not clear that it was a serious problem at the time, but it's become an industry that's clearly causing far more problems than it solves.
ICANN keeping their fee is at least a good start - in theory the registry should also get some cut of the restocking fee as well. In practice this isn't enough money to kill off the whole industry, but it's a least going to cut down on some of the noise.
Yes, you need to be maneuverable if you're in a short-range fight, but this isn't the kind of weapon you use to fight against kamikaze Zodiacs or nearby Soviet ships. But it's presumably not the only weapon you'd put on a ship this size; you'd keep smaller guns or torpedoes or whatever around for occasions when somebody objects to you using long-range artillery.
Personal-firewall software running in Windows isn't perfect, and it's running on Windows, so if something else breaks Windows then it can break through the firewall. They're still not a bad thing (except as you noted, teaching users to always hit "ok".) And one long-term usefulness is that they do make it harder to install some kinds of malware that you might do by naively clicking some web link.
More important, though, is having an external firewall that keeps the riff-raff attacks off your computer, at least long enough for you to download Windows updates and anti-virus updates to a new computer. Typical Windows XP computers without them tend to get owned before they have time to get their updates in place, and by keeping out some of the noise, they also reduce load on the computer's operating system. While NAT is overall an abomination that breaks the Internet End-to-End Model, it does at least block some kinds of attacks, and makes it harder for computers that do get owned to send out packets with forged source addresses.
Too many people find that last step too difficult...
I've been using Kaspersky as my anti-virus, and while it's usually rated as one of the most effective, it's gotten really annoying. At first it was just the hundreds of megs of log-files, though I've mostly limited those. But some time in the last six months, its virus tables added some pattern that's in most of my Eudora mailfile backups, and it'll tell me file names but not position in the file. AFAICT, I received some email that either contained a virus or contained a string that Kaspersky thinks looks like a virus signature, and either it's still there or (more likely) I deleted it but hadn't compressed the mailbox file to get rid of it at the time I did the backups. I've cleaned up a few instances of this, but there's some I just can't find sigh.
The tools that get used to investigate your disk drives are going to depend a lot on who's trying to investigate you and your computers - how much are they willing to spend.
Law enforcement organizations aren't going to waterboard you, which would be against the law, though they might have fun tasing you. And courts have simpler methods - they issue you a subpoena that says to turn over any information you've got, and can make you sit in jail or pay heavy fines for not handing it over, or if it's a civil lawsuit they can decide that you're acting in bad faith and decide in favor of your opponent and make you pay their attorney's costs.
Law enforcement organizations are also highly unlikely to get out the electron microscopes and look for fuzzy bits around the edges of your disk tracks; that's more of an NSA/CIA spy-vs-spy kind of threat model. On the other hand, they are often willing to have some sleep-deprived technician who likes bright lights and loud obnoxious music do the kind of disk recovery that looks at your file systems for the data sitting around in unerased blocks or marked deleted in directory listings.
Fundamentally, if you're storing data on a computer that you don't want anybody else to recover, you need to store it in encrypted form so the only thing that can be recovered is the cyphertext.
For most people, though, the real threat model is that Murphy and BillG gang up on you. For that you need backups, and you need to periodically make sure you can recover your backups, and every couple of years you need to copy the data from old media to new media because otherwise your only copy will be on a 9-track tape or MFM disk. And BillG's still going to make sure that you can't read that proprietary file format that was used by some word processor in 1994. And your corporate IT staff are going to write a backup script that only copies files in Microsoft Office formats, which don't include the.txt and.html you saved them in to prevent that problem. (And yes, that's happened to me during a laptop upgrade, and of course they returned the old drive for credit before they gave the new laptop back to me.)
Fortunately, storage costs have been dropping much faster than Moore's Law predicts, so in theory it's getting easier or at least cheaper to do backups. In practice, Murphy's taken out one of my new 500GB drives, and Maxtor's turned the other one from 500GB into 128/137 GB because the old Maxtor USB-drive case didn't know if the new Maxtor drive supported 48-bit addressing....
Back during the Reagan Administration, when I was working as a tool of the military-industrial complex (:-), we had a VAX lab that we used for classified projects. The Army's rules for wiping disks before declassifying them said that you could either use NSA-approved software (didn't want to do the paperwork to find out if any of that was supported on our Unix versions), an NSA-approved Big Degaussing Magnet (not near *my* lab, thank you!), or physical destruction (yee-hah!)
Our disk drives were RM-05s, which had stacks of a dozen or so 14" platters. Most computer administrators had one on their wall showing the effects of a head crash, with various tracks scraped into the oxide finish. I was no longer running the lab when we decommissioned the VAX, but my successor got to take the disks down to the machine shop in the basement to have them sandblasted. The platter on her wall didn't have any oxide left - it was smooth and shiny metal.
According to Wikipedia, the Japanese did fight the Soviets early in WW II, mainly 1939, around Khalkhin Gol near the Mongolia/Manchukuo border. They lost badly to Soviet General Zhukov's tanks, and did a ceasefire around the time Germany invaded Poland. They didn't fight again until 1945 after the defeat of Germany.
A bunch of the original founders of Havenco got voted off the island years ago, Havenco eventually collapsed without having built much of the infrastructure it had planned to, and Ryan gave a talk about it at some Defcon, and has been doing telecoms in Iraq for a few years now. Since then, Sealand has become less friendly to media piracy.
The right country to host in depends a lot on who you're most likely to annoy. If you're really likely to seriously annoy people, host your website in two countries (plus keep a backup), so that if one website gets shut down you've got your mirror site. It's not a bad idea to keep your DNS server in a different country than your hosting, or at least follow the standard DNS advice about keeping your nameservers on different subnets, which in your case means separate countries.
If you're doing something political, your home country is the most likely to get annoyed at you, and in general if there's a government that's annoyed at your website you'd rather not have it be your home jurisdiction, since you don't want to get arrested or have your bank account seized. If you host in another country, it's harder to defend your website, but that government will have a harder time bothering you personally, and you'll have a backup copy handy.
If you're going to annoy somebody in Country X, hosting in Country Y might be a good idea. Sometimes it's convenient to host in a country that doesn't primarily speak Country X's language. (Everybody speaks some English, but they may be better about having an automated website-setup site in English than actually responding to legal complaints in English. It'll be harder for you to argue with them, but less necessary.) If you're likely to libel somebody, don't host in Britain or Australia; libel laws there are plaintiff-friendly. If you're going to annoy Scientologists, Germany's not very friendly to them. If you're going to annoy US Intellectual Property Owners, you might try China or Russia, but you might end up paying more there. The Caribbean's often friendly, but bandwidth there tends to be overpriced.
Ever notice how Slashdot articles on abuse of the DNS system or attempts to stop it are often on websites that serve the DNS abuser business? I can't tell for sure about domaintools.com, but it looks like they're mainly in business to support people who rent domain names for their advertising value rather than for their ability to indicate the content on a real site. It's possible that I'm mis-characterizing them, but their domain name sale and auction tools and a number of their blog articles look like they're serving the parasite market (just not necessarily the domain-taster part of it.)
Some registrars apparently benefit from it, probably from domain parking services that they provide along with the names or whatever other fees they charge in addition to the $6 ICANN price. Otherwise there wouldn't be so many of them doing it, since it costs them not only the relatively minimal cost of handling transactions but also the cost of capital on the $6/year for the domains they're kiting (though obviously that's only 5/365th of the cost each time, so it's still not much.)
It's easy to remedy it - ICANN just has to stop requiring that the Registry handle those transactions for free; any non-zero cost will help the problem.
And yeah, it's in Google's self-interest. They still get the advertising revenue if anybody's using their ads, they just don't pay the page-owners on it, so legitimate domain name users won't decrease their revenues much, and there'll be fewer content-free pages attracting suckers, so the pages that Google serves up are more likely to get actual useful results. It's possible that they'll lose a bit of revenue, but it's really bottom-feeders.
Another way it's in Google's self-interest is that by reducing the number of content-free domains on the web, there's less trash for their web-search service to crawl through and index.
Google makes money from advertising by charging you money for showing ads, at least if people click on them.
Google shows ads on pages that include Google Adsense banners, so if you can get somebody to look at your web page and click on the ads, Google pays you. Domainer Parasites do this by buying or tasting plausible-sounding names and putting up ad banner pages, usually with no real content, and domain parking services will do the work of implementing them if you don't want to serve them yourself. Domain Tasters kite names by registering them and then returning them during the 5-day grace period, so that they get to collect ad revenue on their parked pages without having even paid for the page, though if the page is going to pay more than the $6/year domain name fee they'll generally buy it and keep it.
If Google stops paying for kited pages, domain taster parasites are going to have to use other advertising services, which are often less convenient or don't pay as much. A big difference is that Google's advertising is usually topical enough that people are more likely to click through, as opposed to more random. Once they've had the domain long enough to actually pay for it, they can switch to Google, but at least it'll cut down on some of the kiting.
One of the requirements for patenting something is that it can't be something obvious to the skilled practitioner. Most of the features he talked about were certainly obvious not only to skilled practitioners in the 1997-2000 timeframe, but to non-technical marketing people, the Nokia smartphone was out, and friends of mine were working on startups in that space, where the major non-obvious detail was "How do we describe this in a way that a VC will give us $4 million after we shake a tree in Menlo Park?", which, alas, we didn't quite find a satisfactory answer to:-)
But these trolls are describing a phone that not only had *each* of the features they claim, but in fact had *all* of them and still fit in a hand-held form factor. I'm pretty skeptical about the ability to fit a GPS device into a phone back in 2000 and still have it be hand-held, though hand-held GPS was certainly available. (I'm even more skeptical about the ability to have a satisfactory battery life if you did.) And I'm even more skeptical about the claim that they actually *did* reduce their concept of these devices to practice.
If you're not required to actually come up with the technology to build the thing, I'm perfectly able to imagine one of these things that fits in your ear canal and runs all year without recharging... So send me my check once you've built the thing.
Also, look at what that 40 GB/month cap costs them - while it's hard to tell exactly what Internet bandwidth costs, since it's been in free-fall for a decade or more, and it's extremely dependent on conditions like directionality, free peering vs. paid transit, a rough ballpark figure is $10/Mbps/month for paid transit 95%ile usage. So that 123kbps is potentially costing them about $1.23. Even that's highly debatable - if most of it's BitTorrent, then potentially 90% of it comes from other customers of the same ISP, so the 123kbps is more like $0.12/mo.
But let's talk about peering vs. transit for a moment - there are roughly three kinds of ISP traffic aggregates out there - content providers, consumer-eyeball providers, and balanced traffic carriers. Balanced carriers will peer with each other for free if they're both big enough that it makes sense, because it lets them offload traffic that they'd otherwise have to carry, and they want to charge money to the content and eyeball providers who have unbalanced traffic. But the content providers and the eyeball providers both want to peer with each other for free, because it lets them serve their highly-complementary customer bases cheaply, and only pay the transit providers to balance geographical and size issues. So BigWebSite and BigCableCo will probably connect for free in San Francisco and New York, but if the BigWebSite doesn't have a server-mirror in Boston, then either they'll use internet transit providers to get that content to consumers in Boston, or else one side will pay the other to cover the transit (e.g. BostonCableCo might need to pay YouTube, but SiliconValleyStartup might need to pay Comcast), and at some point the content provider may want to reduce costs by locating a server in Boston, or paying a caching service like Akamai or AT&T to host one for them. On the other hand, a non-consumer-content user would still need to pay a transit provider, e.g. RandomWidgetFactory probably wouldn't get free service from Comcast because they're not entertaining couch potatoes, and they might be consuming as much inbound bandwidth from their employees browsing the web as their outbound website traffic to potential widget buyers.
So that cable/DSL company probably isn't paying close to market-rates for bandwidth, because they're getting it from the big content providers for free (well, for the cost of interconnect hardware, plus some rack space and GigE cross-connect at Equinix, but that's still in the $1/Mbps/mo range for big users.)
And yeah, the industry's still going to find ways to haul fatter pipes to the chokepoints, and the costs of doing so keep going down, but at least until people are receiving more bits on their computers than on their televisions, multicast solutions of some sort are going to be big players. One thing that helps is that dumb televisions are getting augmented by Tivo and other PVRs, so the video-on-demand business can still use broadcast/multicast to ship the bits to your house, and you'll watch them a bit later. I've certainly stopped watching live TV except for rare random events (or when there are two programs I want at the same time so I'll watch one and Tivo the other) or maybe the evening news; otherwise you've got to wait for the commercials to play.
I get annoyed when people complain about the alleged "duopoly" in US broadband services. From a perspective of how much bandwidth you can get to your house, yeah, there are only two sets of wires, but bandwidth caps, policies on running servers at home, sharing your service with multiple PCs, neighbors, or wifi, static IPs, etc. are all functions of the ISP at the head end, not the wire provider.
My ISP is reselling telco DSL, but I've got a static IP address, no bandwidth caps, no Port 25 blocking, and I can share it with whoever I like. Their policy on running servers and file sharing at home is that they're not selling me walled-garden couch-potato service, they're selling me Internet Connectitivity, and it's up to me to do something cool and worthwhile with it if I want. And there are lots of other ISPs like them - Speakeasy's one of the best-known, and there are also other ISPs that resell telco DSL who do offer "features" like bandwidth caps, URL-censorship to protect your kids, and walled-garden content if you want that sort of thing. I do pay a bit more than I would directly from my telco, but it's in the same general range as other ISPs' static-IP prices. When I first signed up with them, they were doing cool stuff with 802.11 roof-top networks as well, but that hasn't gotten to my town yet.
From an ISO Protocol Stack perspective, who's providing what layers in the US (YMMV in other countries)?
Layer 1 - the telco is providing dry copper wire from the CO to your house
Layer1/Layer 2 - The DSLAM might be provided by the telco, or might be provided by an ILEC such as Covad, New Edge, Megapath, etc.
Layer 2 - Usually DSL runs ATM between your home router and the DSLAM, though sometimes it's frame, and sometimes they wrap a PPPoE layer on top of it to annoy the users and make it easier to cut off people who don't pay their bills. There's usually some regional ATM concentrator network between the DSLAMs and the routers, though sometimes the routers will be colocated with the DSLAMs and the concentration will happen over an IP network instead.
Layer 3 - Usually the router behind the DSLAMs is run by the ISP, so features like DHCP, Static-vs-Dynamic IP addresses, etc. happen here. Sometimes this layer is outsourced - see Layer 8 - or there's some kind of PPPoE tunnel that gets you to the ISP's bigger routers.
More Layer 3 - Once you're at the router, there's some kind of connectivity that gets you to the Internet Backbone. The ARPAnet/NSFnet are long gone these days, and in the US the "backbone" is really about 25 Tier 1 ISPs that interconnect with each other, plus some exchange points where they and other ISPs connect together, so this may mean either your ISP's backbone if they're big or their upstream provider's backbone. In Europe, the exchange points like LINX and AMSIX drive much more of the connectivity, and there's less backbone dependence.
Layers 4-7 - Many ISPs provide applications like email, web servers, IM servers, etc.
Layer 8 - The old IETF T-shirt describes Layer 8 as "Financial" and Layer 9 as "Political". There are some ISPs that provide a layer 4-7 user experience (e.g. email service and login authentication) and do everything else at Layer 8 (outsourcing to a real provider.) That's not as trivial as it sounds - there's a reasonable-sized business market for aggregators who can go out to telcos, cable providers, Wifi Hotspots, etc. around the US or world and sell you one service that works everywhere. And Layer 8 is where issues like bandwidth caps happen.
Layer 9 - Political - All the Network Neutrality, No-servers-at-Home, SMTP-users-must-be-spammers, etc. really lives here, though it may get implemented by mechanisms at lower layers.
Scaling issues are a bit different for a telco DSL service than for a cable network, but they're not _that_ different. A US telco central office typically handles 10K-100K users; a city-wide cable network is similar. Older DSL technology ran the wire back to a DSLAM at the CO, while the newer technology often puts DSLAMs in the green wiring boxes that handle a few blocks so the copper's shorter, and cable equipment typically has a similar-sized box handling 500-1000 users. I don't know which approach FIOS takes. The copper from a new DSLAM to a home can handle about 20-25Mbps, which is about 15Mbps for Television (one approach uses 9 Mbps for one HD channel and 2 Mbps each for three SD channels), and the leftovers are available for Internet data (though some telcos insist on selling it as 1.5/3/6 instead of "whatever's not busy with television right now", which IMHO as a geek is a lame idea about revenue protection.)
In that telco example, the bandwidth limits mean that traffic gets effectively unicast from the DSLAM to the user, because you can't fit 1000 channels of broadcast into 25 Mbps. (By "effectively unicast", I mean it's either actually regular unicast, or it's multicast with only the channels you need on the wire. Same bandwidth etc., just a difference in whether you're in Class D IP address space with multicast handshaking or whether that's all hidden from the home router.) On the other hand, if everybody's watching TV at once, 10K-100K houses at 15 Mbps is 150-1500 Gbps, which isn't realistic. If you feed the CO with multicast, then a GigEthernet can handle about 200 channels of HD or 500 channels of SD, or an OC48 can handle both, and farm it out to everybody who's watching. That's one of the reasons that the telcos want to sell TV as a competing-with-cable service, as opposed to just providing pure transport. (Another is the usual money, competition, etc.)
If everybody's doing typical Internet usage, there are a couple of reasons that the network doesn't melt. The big one is that not everybody's actually burning high bandwidth at once - most of the time you're looking at web pages, maybe pictures, and occasional videos (Youtube etc.), but in practice you can oversubscribe by more than 10:1. Another reason is that TCP reacts to congestion by adjusting transmission speeds and window sizes, so if there are too many people watching Youtube at once, everybody's downloads slow down a bit, but unlike live TV, Youtube doesn't care much.
The other way to get enough bandwidth to the CO is to cache a lot of popular video material there - so either the Akamai model (which is driven by the content providers) or transparent caching run by the ISP (a much older model) can do some of it. It won't catch everything, but I'd hope you could cut Youtube bandwidth demand in half that way.
Disclaimer: This doesn't even *pretend* to be my employer's opinion.
Slashdot Mirror
Herds of tiny cat robots? Cool - that'll go well with the Evil Laugh and the Monocle!
The real reason the wall worked that long is that none of the tigers had previously felt motivated enough to jump at it. Apparently Siberians are more aggressive than Bengals, and maybe the two drunk kids pissed her off or just acted enough like prey or cat toys that she went for them. My cats sit on the couch looking out the window at Bird TV, and when one of them sees a laser pointer red dot he has to jump for it without thinking about it first (the other one says "Hey, stop wavin' that thing around".) And I've seen zoo leopards looking at the crowds, intensely tracking the smaller ones that get separated a bit from their herds; I'd feel much safer around tigers.
The number of tigers in zoos is about 4000.
As many as 3000 tigers may be in farms in China, being raised to sell as traditional medicine for people whose penises aren't big enough or who think their bones will make them stronger.
The number of tigers that are kept as pets by Americans is about 6000. There are animal activists like Tippi Hedren trying to make laws against keeping tigers as pets, because almost nobody who has pet tigers has enough space and resources to let them live like tigers need to, especially the occasional drug dealer in some apartment building in New York who wanted to out-macho his competitors' pit bulls. She's well-intentioned, but the species needs all the genetic diversity it can get, even though tigers aren't meant to live like house-cats.
Or as a friend of mine commented, "If they were six-foot cuddly bunny-rabbits, we'd have called them bunny-rabbits, not tigers!"
Also, if you look at how internet transmission works, while you obviously want geographical redundancy, that doesn't mean that you don't send traffic on all available routes. Carriers are going to make sure they've got enough redundancy for their critical load levels (e.g. the voice network and private-line customers), but if they're doing redundancy at Layer 3 they're going to send traffic across multiple routes because it doesn't make sense to leave them idle.
To some extent, if you're doing Dense Wavelength Division Multiplexing, and if you haven't lit up all your wavelengths (because the optics and routers at the end are expensive), you can sometimes divert some wavelengths to the alternate routes. For instance, you'd provision wavelengths A, B, and C on the west route and Z, Y, X on the east route, and if something breaks you can push them onto the other route. But once your cable fills up, you've got less ability to do that until you build more cables or put even more expensive optics on the ends to light them.
And sometimes you just get surprised - like the big Taiwan earthquake last year that took out N-1 of the undersea cables between northern and southern Asia, which almost all go between Taiwan and the Philippines since that's what the ocean floor shape makes you do. They were spread out far enough to avoid problems with ship-anchors, but the quake was over a wide area. And there was a quake in the Med a couple of years ago that took out more than two cables as well.
Re: your last point, remember that the WTC was attacked twice, the first time with a truck-bomb in 1993, which wasn't successful in taking down the building (because some fscking rude New Yorker had taken the parking place next to the pillar they'd planned to park the truck in
The reason that most of them do it is ad-banner revenue - if enough people hit your web site by typing keywords into their browser that you can pay for it with ad banners or Google adwords or whatever, then you're making a profit. Domain tasting lets you try out a name for nearly-free and see if it generates enough hits over a couple of days to make it worth keeping. Depending on how much revenue you get, it *might* be worth paying the whole $6/year to keep it, or it might only be worth the cost of kiting it, which is still non-zero (it's your cost of money on $6/year), and which risks somebody else grabbing it if their automated tools are faster than yours.
Grabbing names that somebody else is checking the ownership on does two things - one is that it's sometimes possible to make money by scalping the name to them, but also it means that the name had *some* probability of being something that humans might type into a browser, so it's more likely to be worth grabbing for the advertising revenue. It's especially common if a domain name has just expired, because there's a good chance that the site had some traffic on it that you can capture for your own ad banners. A friend of mine just had her domain name stolen after owning it for 10 years - hopefully she can get it back but at least as of last week it was pointing to a domain-parker page.
There's a whole parasite-infested industry called "domainers", who are people who buy up domain names for the advertising revenue and for the possibility of making a profit by selling them either to other domainers or to businesses that actually want the name. They're parasites because they're using name space without providing any useful content, and in many cases they're also filling up Google with pages pointing to their pages, so that Google searches are more likely to get their sites than the interesting stuff you were looking for.
The original domain name grace period rule was ostensibly to prevent people from complaining that they'd paid $35 for a domain name that they'd mistyped or that somebody else claimed was their trademark, and it avoided trademark disputes because you could simply return the name rather than having to sell it to the alleged trademark owner who might then complain that you were trying to name-squat them. It's not clear that it was a serious problem at the time, but it's become an industry that's clearly causing far more problems than it solves.
ICANN keeping their fee is at least a good start - in theory the registry should also get some cut of the restocking fee as well. In practice this isn't enough money to kill off the whole industry, but it's a least going to cut down on some of the noise.
Yes, you need to be maneuverable if you're in a short-range fight, but this isn't the kind of weapon you use to fight against kamikaze Zodiacs or nearby Soviet ships. But it's presumably not the only weapon you'd put on a ship this size; you'd keep smaller guns or torpedoes or whatever around for occasions when somebody objects to you using long-range artillery.
More important, though, is having an external firewall that keeps the riff-raff attacks off your computer, at least long enough for you to download Windows updates and anti-virus updates to a new computer. Typical Windows XP computers without them tend to get owned before they have time to get their updates in place, and by keeping out some of the noise, they also reduce load on the computer's operating system. While NAT is overall an abomination that breaks the Internet End-to-End Model, it does at least block some kinds of attacks, and makes it harder for computers that do get owned to send out packets with forged source addresses.
I've been using Kaspersky as my anti-virus, and while it's usually rated as one of the most effective, it's gotten really annoying. At first it was just the hundreds of megs of log-files, though I've mostly limited those. But some time in the last six months, its virus tables added some pattern that's in most of my Eudora mailfile backups, and it'll tell me file names but not position in the file. AFAICT, I received some email that either contained a virus or contained a string that Kaspersky thinks looks like a virus signature, and either it's still there or (more likely) I deleted it but hadn't compressed the mailbox file to get rid of it at the time I did the backups. I've cleaned up a few instances of this, but there's some I just can't find sigh.
Law enforcement organizations aren't going to waterboard you, which would be against the law, though they might have fun tasing you. And courts have simpler methods - they issue you a subpoena that says to turn over any information you've got, and can make you sit in jail or pay heavy fines for not handing it over, or if it's a civil lawsuit they can decide that you're acting in bad faith and decide in favor of your opponent and make you pay their attorney's costs.
Law enforcement organizations are also highly unlikely to get out the electron microscopes and look for fuzzy bits around the edges of your disk tracks; that's more of an NSA/CIA spy-vs-spy kind of threat model. On the other hand, they are often willing to have some sleep-deprived technician who likes bright lights and loud obnoxious music do the kind of disk recovery that looks at your file systems for the data sitting around in unerased blocks or marked deleted in directory listings.
Fundamentally, if you're storing data on a computer that you don't want anybody else to recover, you need to store it in encrypted form so the only thing that can be recovered is the cyphertext.
For most people, though, the real threat model is that Murphy and BillG gang up on you. For that you need backups, and you need to periodically make sure you can recover your backups, and every couple of years you need to copy the data from old media to new media because otherwise your only copy will be on a 9-track tape or MFM disk. And BillG's still going to make sure that you can't read that proprietary file format that was used by some word processor in 1994. And your corporate IT staff are going to write a backup script that only copies files in Microsoft Office formats, which don't include the
Fortunately, storage costs have been dropping much faster than Moore's Law predicts, so in theory it's getting easier or at least cheaper to do backups. In practice, Murphy's taken out one of my new 500GB drives, and Maxtor's turned the other one from 500GB into 128/137 GB because the old Maxtor USB-drive case didn't know if the new Maxtor drive supported 48-bit addressing....
Our disk drives were RM-05s, which had stacks of a dozen or so 14" platters. Most computer administrators had one on their wall showing the effects of a head crash, with various tracks scraped into the oxide finish. I was no longer running the lab when we decommissioned the VAX, but my successor got to take the disks down to the machine shop in the basement to have them sandblasted. The platter on her wall didn't have any oxide left - it was smooth and shiny metal.
According to Wikipedia, the Japanese did fight the Soviets early in WW II, mainly 1939, around Khalkhin Gol near the Mongolia/Manchukuo border. They lost badly to Soviet General Zhukov's tanks, and did a ceasefire around the time Germany invaded Poland. They didn't fight again until 1945 after the defeat of Germany.
A bunch of the original founders of Havenco got voted off the island years ago, Havenco eventually collapsed without having built much of the infrastructure it had planned to, and Ryan gave a talk about it at some Defcon, and has been doing telecoms in Iraq for a few years now. Since then, Sealand has become less friendly to media piracy.
If you're doing something political, your home country is the most likely to get annoyed at you, and in general if there's a government that's annoyed at your website you'd rather not have it be your home jurisdiction, since you don't want to get arrested or have your bank account seized. If you host in another country, it's harder to defend your website, but that government will have a harder time bothering you personally, and you'll have a backup copy handy.
If you're going to annoy somebody in Country X, hosting in Country Y might be a good idea. Sometimes it's convenient to host in a country that doesn't primarily speak Country X's language. (Everybody speaks some English, but they may be better about having an automated website-setup site in English than actually responding to legal complaints in English. It'll be harder for you to argue with them, but less necessary.)
If you're likely to libel somebody, don't host in Britain or Australia; libel laws there are plaintiff-friendly. If you're going to annoy Scientologists, Germany's not very friendly to them. If you're going to annoy US Intellectual Property Owners, you might try China or Russia, but you might end up paying more there. The Caribbean's often friendly, but bandwidth there tends to be overpriced.
Ever notice how Slashdot articles on abuse of the DNS system or attempts to stop it are often on websites that serve the DNS abuser business? I can't tell for sure about domaintools.com, but it looks like they're mainly in business to support people who rent domain names for their advertising value rather than for their ability to indicate the content on a real site. It's possible that I'm mis-characterizing them, but their domain name sale and auction tools and a number of their blog articles look like they're serving the parasite market (just not necessarily the domain-taster part of it.)
It's easy to remedy it - ICANN just has to stop requiring that the Registry handle those transactions for free; any non-zero cost will help the problem.
And yeah, it's in Google's self-interest. They still get the advertising revenue if anybody's using their ads, they just don't pay the page-owners on it, so legitimate domain name users won't decrease their revenues much, and there'll be fewer content-free pages attracting suckers, so the pages that Google serves up are more likely to get actual useful results. It's possible that they'll lose a bit of revenue, but it's really bottom-feeders.
Another way it's in Google's self-interest is that by reducing the number of content-free domains on the web, there's less trash for their web-search service to crawl through and index.
Google shows ads on pages that include Google Adsense banners, so if you can get somebody to look at your web page and click on the ads, Google pays you. Domainer Parasites do this by buying or tasting plausible-sounding names and putting up ad banner pages, usually with no real content, and domain parking services will do the work of implementing them if you don't want to serve them yourself. Domain Tasters kite names by registering them and then returning them during the 5-day grace period, so that they get to collect ad revenue on their parked pages without having even paid for the page, though if the page is going to pay more than the $6/year domain name fee they'll generally buy it and keep it.
If Google stops paying for kited pages, domain taster parasites are going to have to use other advertising services, which are often less convenient or don't pay as much. A big difference is that Google's advertising is usually topical enough that people are more likely to click through, as opposed to more random. Once they've had the domain long enough to actually pay for it, they can switch to Google, but at least it'll cut down on some of the kiting.
But these trolls are describing a phone that not only had *each* of the features they claim, but in fact had *all* of them and still fit in a hand-held form factor. I'm pretty skeptical about the ability to fit a GPS device into a phone back in 2000 and still have it be hand-held, though hand-held GPS was certainly available. (I'm even more skeptical about the ability to have a satisfactory battery life if you did.) And I'm even more skeptical about the claim that they actually *did* reduce their concept of these devices to practice.
If you're not required to actually come up with the technology to build the thing, I'm perfectly able to imagine one of these things that fits in your ear canal and runs all year without recharging... So send me my check once you've built the thing.
But let's talk about peering vs. transit for a moment - there are roughly three kinds of ISP traffic aggregates out there - content providers, consumer-eyeball providers, and balanced traffic carriers. Balanced carriers will peer with each other for free if they're both big enough that it makes sense, because it lets them offload traffic that they'd otherwise have to carry, and they want to charge money to the content and eyeball providers who have unbalanced traffic. But the content providers and the eyeball providers both want to peer with each other for free, because it lets them serve their highly-complementary customer bases cheaply, and only pay the transit providers to balance geographical and size issues. So BigWebSite and BigCableCo will probably connect for free in San Francisco and New York, but if the BigWebSite doesn't have a server-mirror in Boston, then either they'll use internet transit providers to get that content to consumers in Boston, or else one side will pay the other to cover the transit (e.g. BostonCableCo might need to pay YouTube, but SiliconValleyStartup might need to pay Comcast), and at some point the content provider may want to reduce costs by locating a server in Boston, or paying a caching service like Akamai or AT&T to host one for them. On the other hand, a non-consumer-content user would still need to pay a transit provider, e.g. RandomWidgetFactory probably wouldn't get free service from Comcast because they're not entertaining couch potatoes, and they might be consuming as much inbound bandwidth from their employees browsing the web as their outbound website traffic to potential widget buyers.
So that cable/DSL company probably isn't paying close to market-rates for bandwidth, because they're getting it from the big content providers for free (well, for the cost of interconnect hardware, plus some rack space and GigE cross-connect at Equinix, but that's still in the $1/Mbps/mo range for big users.)
And yeah, the industry's still going to find ways to haul fatter pipes to the chokepoints, and the costs of doing so keep going down, but at least until people are receiving more bits on their computers than on their televisions, multicast solutions of some sort are going to be big players. One thing that helps is that dumb televisions are getting augmented by Tivo and other PVRs, so the video-on-demand business can still use broadcast/multicast to ship the bits to your house, and you'll watch them a bit later. I've certainly stopped watching live TV except for rare random events (or when there are two programs I want at the same time so I'll watch one and Tivo the other) or maybe the evening news; otherwise you've got to wait for the commercials to play.
My ISP is reselling telco DSL, but I've got a static IP address, no bandwidth caps, no Port 25 blocking, and I can share it with whoever I like. Their policy on running servers and file sharing at home is that they're not selling me walled-garden couch-potato service, they're selling me Internet Connectitivity, and it's up to me to do something cool and worthwhile with it if I want. And there are lots of other ISPs like them - Speakeasy's one of the best-known, and there are also other ISPs that resell telco DSL who do offer "features" like bandwidth caps, URL-censorship to protect your kids, and walled-garden content if you want that sort of thing. I do pay a bit more than I would directly from my telco, but it's in the same general range as other ISPs' static-IP prices. When I first signed up with them, they were doing cool stuff with 802.11 roof-top networks as well, but that hasn't gotten to my town yet.
From an ISO Protocol Stack perspective, who's providing what layers in the US (YMMV in other countries)?
In that telco example, the bandwidth limits mean that traffic gets effectively unicast from the DSLAM to the user, because you can't fit 1000 channels of broadcast into 25 Mbps. (By "effectively unicast", I mean it's either actually regular unicast, or it's multicast with only the channels you need on the wire. Same bandwidth etc., just a difference in whether you're in Class D IP address space with multicast handshaking or whether that's all hidden from the home router.) On the other hand, if everybody's watching TV at once, 10K-100K houses at 15 Mbps is 150-1500 Gbps, which isn't realistic. If you feed the CO with multicast, then a GigEthernet can handle about 200 channels of HD or 500 channels of SD, or an OC48 can handle both, and farm it out to everybody who's watching. That's one of the reasons that the telcos want to sell TV as a competing-with-cable service, as opposed to just providing pure transport. (Another is the usual money, competition, etc.)
If everybody's doing typical Internet usage, there are a couple of reasons that the network doesn't melt. The big one is that not everybody's actually burning high bandwidth at once - most of the time you're looking at web pages, maybe pictures, and occasional videos (Youtube etc.), but in practice you can oversubscribe by more than 10:1. Another reason is that TCP reacts to congestion by adjusting transmission speeds and window sizes, so if there are too many people watching Youtube at once, everybody's downloads slow down a bit, but unlike live TV, Youtube doesn't care much.
The other way to get enough bandwidth to the CO is to cache a lot of popular video material there - so either the Akamai model (which is driven by the content providers) or transparent caching run by the ISP (a much older model) can do some of it. It won't catch everything, but I'd hope you could cut Youtube bandwidth demand in half that way.
Disclaimer: This doesn't even *pretend* to be my employer's opinion.
The author makes some good points. I'm not saying that it isn't flamebait, but it deserves to be +3 Flamebait, not +0 Flamebait :-)
The Jeff Goldblum version, not the Vincent Price version...