Sure, there's some correlation, because advertisers pay you more if they think you're giving them more viewers. But the viewers that matter least are the ones who won't spend $100-200 for a D/A converter box or can't talk their parents into spending it. Additionally, in most markets, there's a transition option, which is to kill off the UHF analog stations first and leave VHF running on analog, or the other way around if the networks can't get the FCC to give them UHF digital bandwidth as a replacement.
This is Long Island, not just Brooklyn and Queens. It's like stretching Overland Park out to Lawrence and Shawnee Mission. Doesn't mean you'll be able to get it anywhere west of Wichita or up in Atchison any time soon, but if you're living out in farm country you're obviously there for the peace and quiet.
50 Mbps symmetric service isn't something for the average home user, who can be served much less expensively with asymmetric services (even if they're 50/5 or whatever.) This is a business service, presumably with a business price level, and since it's symmetric and fast they're presumably permitting businesses to run web servers (just like they *should* be doing with consumer service but are too stupid and stubborn to do.)
The technically cool thing you get from this kind of service is that the physical cable routes are generally different from telco cable routes, so not only can a business buy lots of bandwidth for a reasonable or low price, but it provides backhoe protection for their telco-based services without having to go to the expense of building physically diverse fiber rings.
They're going to allow high-speed symmetric connections because there's demand from customers who are willing to pay more for it. The press release doesn't talk about pricing, but this isn't going to be the same $20-50/month that typical consumer cable modems cost, and probably not even the $50-100 that low-end business cable service costs. I'm guessing it's somewhere in the $500 range, maybe $150-1000.
The interesting issues are going to be pricing, average throughput (e.g. how many people are you sharing your upstream with), and policies about port blocking (they're presumably going to allow web servers, because that's the kind of application that needs 50 Mbps upstreams.
For a business, besides price, the technically cool thing about high-speed cable modem service is that it's not using the same wiring from your office to the telco POP that almost everything else users, so you get some protection from street construction crews and Bubba the Backhoe driver that you'd otherwise only be able to get by buying a higher-end fiber ring service from the telco or using a short-haul wireless connection to a nearby wireless provider. So depending on your price and reliability needs, you can either use this for cheap fast unreliable service, or for cheap reliability improvement to your existing more expensive service, as well as for cheap speed improvements to your regular service. After all, if what you really need is 5-10 Mbps, then getting a 50 Mbps service that's oversubscribed a bit too heavily and priced like a T1 line is almost always a big win.
Repair Speed is the main business problem with cable modem services - the economics of providing $30/month service depend on piggybacking on consumer cable TV service, which means you've got enough technicians and repair trucks to go fix it if it breaks, but if it's Friday night in a bad snowstorm, and your customer's TV service goes out, they can just watch videos or play with their kids or read books until Monday when the snowplows have finished clearing the streets. Low-end "business" cable may mean you get better help-desk service, and maybe the truck goes to your building a bit earlier, but it doesn't put any more trucks on the street. This service may be priced high enough to pay for better service than that.
Commuting on Long Island I used to have a project in Syosset that required me to commute there from central New Jersey for a month. Took about 1.5 hours each way, unless traffic was worse than usual, like the days that it was faster to walk across Staten Island than to drive. Hard drugs would have made driving too difficult, but a Grateful Dead concert tape is about the length of a round-trip, which was at least a good substitute. I tried taking the train one time when it was going to snow heavily - about 2.5 hours to get from Jersey to NYC to the LIRR to whatever the nearest station was, get a taxi to the office, and find out that they were closing because of snow (:-), and the LIRR was far noisier and bouncier than the New Jersey trains so it wasn't possible to do any real work while riding them.
Yes, you're using a consumer service which has cretinously designed policies implemented badly with poor communications to the users so they'll be scared into only doing couch potato stuff instead of participating in the Internet, but you probably knew that when you bought cable modem service (or certainly figured it out quickly soon after:-) But the problems aren't technology - they're policies and terms of service, and it's not that they can't provide higher speeds, they're just not providing them to *you* because they think you're a couch potato who wants to pay $30-40/month for Yet Another flavor of Pay-Per-View while their kids play online games.
This is a "Business Service", designed to give customers business customers different policies at a different price point. That doesn't mean that they'll necessarily implement clueful open-usage policies at a very low price, or even at a medium price, but they're perfectly capable of doing it if they feel like it. (After all, they went to a lot of work to implement the cretinously stupid policies they've imposed on your service, so you know they're at least persistent...) The system is implemented using Hybrid Fiber Coax, so the bandwidth from the cable head end in your neighborhood to their hub sites is as big as they feel like providing (depending on what hardware they're using to light their fibers and whether they own dark fiber or rent bandwidth from telcos, probably the former. It's symmetric service.) The asymmetric part in consumer cable modem service is on the shared cable from the head end to the home; this service appears to get around that by dedicating cable, or else by dedicating bandwidth on the cable - it's a bit hard to tell, but not really relevant. Is there enough upstream bandwidth that your business could really burn a full-speed 50 Mbps in both directions full time? Probably not - if the service costs $10,000/month, I'd find that really unattractively priced, and if it costs $100-500/month and you can figure out what the real performance is, that's really just fine, because paying the price of a dedicated T1 line and occasionally getting up to 30 times that fast is a pretty good deal.
I've seen some cable companies that offer "Business Service" at about twice the price of consumer service - usually it doesn't mean that you're getting real fully-open Internet access, just that the people running their consumer service are greedy morons who block IPSEC and other VPN services because "consumer service isn't for business use" and they think they can get another $30/month in return for unblocking it and giving you higher priority at the help desk. But there are other cable companies that aren't so stupid (they might be just as greedy, they're just a bit more selective in how they rip you off.)
Disclaimer: I'm not currently an employee of a cable modem company, just a stockholder, and yes, they're one of the stupid+greedy types unless they've somehow wised up recently, which is unlikely. Cable modem would be faster than my DSL service, and a bit cheaper, but it's worth paying extra for sonic.net DSL, because they're actually letting me use the Internet service I'm paying for (Speakeasy's another service with similar policies.)
While many of the transatlantic cable systems are near each other, and some of them do share cable heads landing sites, there's also a lot of diversity, put in there specifically to prevent single events from taking out redundant systems all at once, and they're designed in self-repairing rings and meshes for most networks. The Pacific and Caribbean cable systems are pretty much the same way - it takes a lot of time and money to get diversity, and it's done because otherwise you can lose all your connectivity too easily. In India, there are at least three major cable landing locations, and systems like SMW-3 and FLAG use at least two of them, with land and water connections between the landings, to avoid getting disconnected. But Pakistan only has one spur off of SMW-3, and there are other small countries with similar problems along the Persian Gulf.
That doesn't mean you can't have multiple failures that take out redundant systems - about a year ago, there were multiple cable cuts on different sides of Singapore that killed parts of some of the cable systems, so carriers who only used one cable consortium were in trouble for a couple of weeks. Similarly, there was an earthquake in the Mediterranean a couple of years ago that took out parts of half a dozen cable systems, and it took a long time to get them all fixed.
Land-based internet peering points in the US do have the possibility of things going wrong - but that's why any respectably large ISP has physically diverse connections into their important buildings, and access rings using those connections that can restore around failures, and big ISPs peer with each other at multiple locations. There are occasionally geographically entertaining problems, like that railroad tunnel near Baltimore that caught fire a few years back, taking out the circuits from several major ISPs - railroad right-of-way is a very popular way to route long-haul fiber, and often carries multiple long-haul providers as well as local telcos. Fortunately, my employer's network didn't use that tunnel, but we had sufficient diversity in that area that cutting one of our cables would have minimal impact (we design everything with that objective, but there are places like crossing the Rockies where you sometimes have to go a long ways to get an alternate route.
There's been a global cooperative telecom project called TAE for a number of years, which has built fiber networks across much of central Asia, including connecting up to Tehran and other parts of Iran and the whole mess of troubled countries around the Caspian Sea and geographically difficult areas like northwest China. There have been proposals to extend it across northern Pakistan and then to Kabul, but I couldn't find anything newer than 2002 so I don't know if it's being built yet.
But basically if they can run an oil pipeline along many of these regions, they can drag a fiber optic cable along with it; dealing with local telecom bureaucrats is often tougher than installing the cable system across the mountains.
There's some work going on connecting northern Pakistan with nearby parts of India, which is politically significant, just as restarting the bus line was.
There's been a proposal to extend the TAE Trans-Asia-Europe Silk Road cable system from Tehran across northern Pakistan to Islamabad, and from there over to Kabul. The newest documents I could find in Google were from about 2002, and while much of central Asia appears to be connected, I couldn't tell if the Pakistan parts had been completed or not. Also, it's up north, while Karachi is down south and gets the SMW-3 (aka SEA-ME-WE-III) cable, which is the one that's down.
Of course, when you've got telecom monopolies, that seriously degrades your ability to get competitive diverse services, which degrades your ability to create a market that encourages more people to build connectivity. India has theoretically liberalized, but VSNL still seems to have a strong hold on most of the major cable landings, which has been a problem, since there's lots of fiber passing nearby on FLAG, SMW3, etc, and lots of terrestrial fiber to connect it to, if you could just get the stuff up the beach onto dry land without some bureaucrat trying to prevent competition. (Too many US politicians whine about outsourcing - they should only imagine what would have happened if India's telecoms had been liberalized five years earlier and caught more of the 1990s boom.)
The small appliances are cheap, compared to most general-purpose low power x86 servers. It's hard to build a "decent" system for under $150-200, and it's generally going to include a graphics system and want a keyboard and mouse, while you can get a typical appliance for $29-59 including a 4-port or 8-port hub. It's not as flexible, so you probably wouldn't use it for your servers, but it's a good start for protecting client-side users.
Does anybody know if you can boot Knoppix from external DVDs using USB2, assuming your BIOS is new enough to do USB booting at all? My main desktop only has a CD burner right now, and iTunes complains about not being able to read audio from it, so I'm trying to decide whether to buy a cheaper internal DVD burner or a slightly more expensive USB2 version, which would let me use the burner on my laptop which only has DVD-ROM capabilities. (I suppose I could add Firewire to the box as an alternative, but I don't think my BIOS has any chance of booting from that.)
UnionFS gets rave reviews, but is it possible to write the DVD in some extensible format that lets you burn the basic Knoppix onto it and then burn some extra data tracks for other applications (and then mount them with UnionFS or whatever)? That would make it easy to upgrade packages and to add a bunch of user data onto the basic distro, without having to rebuild your own.
Also, I'm curious if the files are compressed? It's a tradeoff of storage and disk read speed vs. CPU speed, and unlike CDs, I'm not sure if it's a win, but DVDs only have ~4.7GB for regular or 8.5 for double-layer, so you'll run out of space after a while:-)
Back in the 80s, I worked for a defense contractor. There are lots of things that lead to the $500 hammer problem
Overspecification - The customer lists *all* the specs they like from *all* the products on the market, even though no single product meets all those specs.
Purchasing Bureaucrat requirements filtering and process formality - The vendor doesn't actually get to talk to the end user who wants the product - the vendor talks to a bunch of purchasing bureaucrats who only have a vague notion of what the end user wants, so they pad some of the requirements to be sure, and the end user has usually padded some of the requirements because he knows that the purchasing bureaucrats don't have a clue what they're doing. And the vendor can't simply sit down and talk to the end user about it, because purchasing rules usually prevent it (to prevent a whole 'nother set of ways that vendors can pad costs and rip off the government.) And usually some PHB in the process read some article in Jet Engine World or Hammer Manufacturer's Digest about some cool new innovation in jet engine hammers and wants to make sure that the end user's don't get ripped off by some vendor selling them an inferior hardware-store hammer.
Cost Accounting for Contract Issues. When you need to buy a $5m jet engine and a $5 jet engine hammer, it's going to cost at least $1000 for the paperwork, and that gets accounted for as part of the items being purchased. So you end up with two line items in the contract, a $5,000,500 jet engine and a $505 hammer.
Fixed vs. Variable Manufacturing Costs and Small Production Runs - Most manufacturing processes have a fixed cost to set up the manufacturing equipment and a variable cost that's the per-unit cost of making the items. So if it costs $5000 to set up the hammer-making equipment and $1 in variable costs for each hammer, Walmart is going to make a million hammers at a cost of $1+(5000/1000000) = $1.0005 each, mark them up to $10, and hold a big price-rollback sale with them marked down to $5 (which includes about a buck a hammer for advertising.) But the Air Force is going to order one mil-spec jet engine hammer with each of the ten jet engines they're buying, so they're going to pay $1 + (5000/10) = $501 per hammer, and then they're going to order a dozen mil-spec helicopter rotor hammers, which are slightly different from jet-engine hammers, so the machines need to be set up again, and they'll pay $1 + (5000/12) = $417.66 per hammer.
Would it be cheaper to order a slightly *better* hammer that can be used for both, at a cost of $2 + (5000/22) = $229/hammer? Yup, but nobody could justify the cost of doing the paperwork to explain why they're buying fancier hammers, leave aside the costs of convening a MIL-SPEC-Joint-Tactical-Hammer Design Committee, or dealing with the auditors who have to explain why Halliburton is importing thousands of $100 MIL-SPEC-Joint-Tactical-Hammer units to Iraq where they're going to use them for busting up sheetrock (which could be done with $5 Walmart hammers) or using them in place of the $500 MIL-Spec-Titanium-Tank-Wheel-Hammer which is made by a contractor in Senator Foghorn's district.
The Bunker is a data center housed in a former missile base in the UK. It's underground, which makes cooling much easier, and has highly redundant commercial power feeds in addition to the stuff they've added themselves. Being able to recycle that kind of space kept their construction costs low and keeps their cooling costs low, and it also looks cool, which has been useful for them to sell service to bankers and other reliability-paranoid types.
I got my license in 1973, when gas prices were going from 29 cents to 45, on their way to the appallingly high price of $1/gallon about 5 years later, and they stayed that high for most of the Reagan/Bush years.
When I moved to the Bay Area in 1993, there was one gas station in Livermore that usually had gas for prices between 0.85 to $1, and prices stayed relatively low for a few years after that in spite of California's increases in gas tax. Then *you* showed up....
I don't have enough WinXP-specific experience to know how many applications actually break when you're running them as non-root, but most of them require your to be admin to install them. One way to do this is to log off from your non-priv account, log on as admin, install the stuff, log off as admin, and log back in as yourself. I normally do that, and it usually works.
Unfortunately, there are a bunch of applications for which this doesn't work right, including iTunes - the first piece of Apple software I've used that didn't "just work". When I installed iTunes, as root, it created an iTunes config for root, but when I logged in as myself, it created a separate iTunes config for me, and I not only had to input lots of long registration numbers again (:-), but the tunes I'd downloaded to root's account aren't accessible from my account and vice versa (or at least, it's well hidden if they are.) Very annoying.
Some things are worse about multiple users - my USB scanner gets hopelessly confused by having multiple people logged in. As far as I can tell, when I first log in as one user, its software scans the USB and finds it, and when I log in as a different user, it does the same thing, except something's locked up to the first person who logged in.
(As somebody else said about their home setup, I've got three accounts on the machine - root, my non-admin account, and my wife's account, which has admin privileges so she can install software and run picky software, and we use fast-user-switching between them.)
I'm not sure if that's always true, but it's certainly true for the VPN software version I use to access the network at work. Very, Very Annoying! It means that I can no longer set up my home PC to access the VPN the way I did when the home machine ran Win98/WinME, since I use XP's fast user switching between root, my non-priv account, and my wife's account.
The difference is that you're setting things up so one badly-written application can run as Admin when the user wants it, but most of your other programs are running as non-privileged-user, vs. setting the user to be Admin for everything they run. Sure, when you're running that one program, you're still exposed, but at least for most of your applications, they're less dangerous.
Re:Unlimited to Handheld != Unlimited to PC card
on
Tetherless Wireless
·
· Score: 1
Depends on your hardware - a lot of the cellphone companies won't sell you the right kinds of equipment that works with their cheaper services and also connects to a PC, using all the usual locked-hardware tricks to preserve it. Very annoying of them:-)
On the other hand, a friend of mine has the Sprint low-cost-unlimited-data service from early on that they've stopped selling (I think he's paying $30/month, maybe less) and it works just fine.
So put your OS and commonly used read-mostly files on the flash, and mount a piece of rotating machinery for/home or whatever, and either don't swap or else swap to the disk drive. If you're paranoid, build a file system that writes to disk in parallel to the flash, so you won't lose anything.
Writing flash may be slow, but reading is fast, and there's no rotational latency to worry about. So put your OS and applications in the flash, the way you would with CD for Knoppix, and that stuff will be a LOT faster. Maybe you use flash for writes, or maybe use a hard drive to mount/usr, or maybe do some sort of translucent file system thing that initially writes to disk and then migrates stuff up to flash.
Yes, writing to flash memory is generally slow. But reading is really fast, and you don't need to wait for rotating machinery to get to the right location to start, so it can really be *much* faster, and you get most of the power and performance savings by speeding up reads. If you're using USB-based flash sticks, obviously you're limited by the quality of the USB drivers, but USB2 itself is faster than most disk busses and isn't a bottleneck, though some people have argued that Firewire is lighter and faster.
Maybe you don't need to use flash for everything, and as you say hard drives are better for repeated use (for *writing*). But a large amount of most people's disk drive usage is reads from installed software or data files they reuse - your machine will be a lot faster and quieter if you're running the OS and common applications from the flash, even if you haven't mounted most of/home on the flash drive. Now that a 1GB USB flash is about $50, it's time to start experimenting with cache strategies - for most Linux applications, that means USB2 flash drives for now. After all, Knoppix fits in ~700MB, and you could use the same compression strategies for a flash drive.
Also, for performance, disk drives have rotational latency, and flash doesn't, so even if reads aren't quite as fast (especially compressed reads that use yor CPU), the effective speed is generally a lot better.
Unlimited to Handheld != Unlimited to PC card
on
Tetherless Wireless
·
· Score: 1
It's nice to get blazingly fast bandwidth to a handheld device, but unless it's got a hard disk bigger than 40GB and you're running BitTorrent, it's fundamentally really low usage for the wireless company. Many of the cellular companies that offer $20 "unlimited" service will not let you use that service to connect your phone/handheld to a PC, or let you use it with a cellular card in your PC, or at least won't let you do it without charging you $80-100. So that phone with a Bluetooth in it isn't allowed to use the Bluetooth to connect to your laptop..
A friend of mine lived in an area that didn't have DSL yet, but he could get the ~56kbps Metricom service, and was the only customer near his nearest Metricom lamppost, so he usually got pretty full bandwidth for his household of four people. Sure beat dialup.
Slashdot Mirror
Sure, there's some correlation, because advertisers pay you more if they think you're giving them more viewers. But the viewers that matter least are the ones who won't spend $100-200 for a D/A converter box or can't talk their parents into spending it. Additionally, in most markets, there's a transition option, which is to kill off the UHF analog stations first and leave VHF running on analog, or the other way around if the networks can't get the FCC to give them UHF digital bandwidth as a replacement.
This is Long Island, not just Brooklyn and Queens. It's like stretching Overland Park out to Lawrence and Shawnee Mission. Doesn't mean you'll be able to get it anywhere west of Wichita or up in Atchison any time soon, but if you're living out in farm country you're obviously there for the peace and quiet.
The technically cool thing you get from this kind of service is that the physical cable routes are generally different from telco cable routes, so not only can a business buy lots of bandwidth for a reasonable or low price, but it provides backhoe protection for their telco-based services without having to go to the expense of building physically diverse fiber rings.
The interesting issues are going to be pricing, average throughput (e.g. how many people are you sharing your upstream with), and policies about port blocking (they're presumably going to allow web servers, because that's the kind of application that needs 50 Mbps upstreams.
For a business, besides price, the technically cool thing about high-speed cable modem service is that it's not using the same wiring from your office to the telco POP that almost everything else users, so you get some protection from street construction crews and Bubba the Backhoe driver that you'd otherwise only be able to get by buying a higher-end fiber ring service from the telco or using a short-haul wireless connection to a nearby wireless provider. So depending on your price and reliability needs, you can either use this for cheap fast unreliable service, or for cheap reliability improvement to your existing more expensive service, as well as for cheap speed improvements to your regular service. After all, if what you really need is 5-10 Mbps, then getting a 50 Mbps service that's oversubscribed a bit too heavily and priced like a T1 line is almost always a big win.
Repair Speed is the main business problem with cable modem services - the economics of providing $30/month service depend on piggybacking on consumer cable TV service, which means you've got enough technicians and repair trucks to go fix it if it breaks, but if it's Friday night in a bad snowstorm, and your customer's TV service goes out, they can just watch videos or play with their kids or read books until Monday when the snowplows have finished clearing the streets. Low-end "business" cable may mean you get better help-desk service, and maybe the truck goes to your building a bit earlier, but it doesn't put any more trucks on the street. This service may be priced high enough to pay for better service than that.
Commuting on Long Island I used to have a project in Syosset that required me to commute there from central New Jersey for a month. Took about 1.5 hours each way, unless traffic was worse than usual, like the days that it was faster to walk across Staten Island than to drive. Hard drugs would have made driving too difficult, but a Grateful Dead concert tape is about the length of a round-trip, which was at least a good substitute. I tried taking the train one time when it was going to snow heavily - about 2.5 hours to get from Jersey to NYC to the LIRR to whatever the nearest station was, get a taxi to the office, and find out that they were closing because of snow (:-), and the LIRR was far noisier and bouncier than the New Jersey trains so it wasn't possible to do any real work while riding them.
This is a "Business Service", designed to give customers business customers different policies at a different price point. That doesn't mean that they'll necessarily implement clueful open-usage policies at a very low price, or even at a medium price, but they're perfectly capable of doing it if they feel like it. (After all, they went to a lot of work to implement the cretinously stupid policies they've imposed on your service, so you know they're at least persistent...) The system is implemented using Hybrid Fiber Coax, so the bandwidth from the cable head end in your neighborhood to their hub sites is as big as they feel like providing (depending on what hardware they're using to light their fibers and whether they own dark fiber or rent bandwidth from telcos, probably the former. It's symmetric service.) The asymmetric part in consumer cable modem service is on the shared cable from the head end to the home; this service appears to get around that by dedicating cable, or else by dedicating bandwidth on the cable - it's a bit hard to tell, but not really relevant. Is there enough upstream bandwidth that your business could really burn a full-speed 50 Mbps in both directions full time? Probably not - if the service costs $10,000/month, I'd find that really unattractively priced, and if it costs $100-500/month and you can figure out what the real performance is, that's really just fine, because paying the price of a dedicated T1 line and occasionally getting up to 30 times that fast is a pretty good deal.
I've seen some cable companies that offer "Business Service" at about twice the price of consumer service - usually it doesn't mean that you're getting real fully-open Internet access, just that the people running their consumer service are greedy morons who block IPSEC and other VPN services because "consumer service isn't for business use" and they think they can get another $30/month in return for unblocking it and giving you higher priority at the help desk. But there are other cable companies that aren't so stupid (they might be just as greedy, they're just a bit more selective in how they rip you off.)
Disclaimer: I'm not currently an employee of a cable modem company, just a stockholder, and yes, they're one of the stupid+greedy types unless they've somehow wised up recently, which is unlikely. Cable modem would be faster than my DSL service, and a bit cheaper, but it's worth paying extra for sonic.net DSL, because they're actually letting me use the Internet service I'm paying for (Speakeasy's another service with similar policies.)
That doesn't mean you can't have multiple failures that take out redundant systems - about a year ago, there were multiple cable cuts on different sides of Singapore that killed parts of some of the cable systems, so carriers who only used one cable consortium were in trouble for a couple of weeks. Similarly, there was an earthquake in the Mediterranean a couple of years ago that took out parts of half a dozen cable systems, and it took a long time to get them all fixed.
Land-based internet peering points in the US do have the possibility of things going wrong - but that's why any respectably large ISP has physically diverse connections into their important buildings, and access rings using those connections that can restore around failures, and big ISPs peer with each other at multiple locations. There are occasionally geographically entertaining problems, like that railroad tunnel near Baltimore that caught fire a few years back, taking out the circuits from several major ISPs - railroad right-of-way is a very popular way to route long-haul fiber, and often carries multiple long-haul providers as well as local telcos. Fortunately, my employer's network didn't use that tunnel, but we had sufficient diversity in that area that cutting one of our cables would have minimal impact (we design everything with that objective, but there are places like crossing the Rockies where you sometimes have to go a long ways to get an alternate route.
But basically if they can run an oil pipeline along many of these regions, they can drag a fiber optic cable along with it; dealing with local telecom bureaucrats is often tougher than installing the cable system across the mountains.
There's some work going on connecting northern Pakistan with nearby parts of India, which is politically significant, just as restarting the bus line was.
Of course, when you've got telecom monopolies, that seriously degrades your ability to get competitive diverse services, which degrades your ability to create a market that encourages more people to build connectivity. India has theoretically liberalized, but VSNL still seems to have a strong hold on most of the major cable landings, which has been a problem, since there's lots of fiber passing nearby on FLAG, SMW3, etc, and lots of terrestrial fiber to connect it to, if you could just get the stuff up the beach onto dry land without some bureaucrat trying to prevent competition. (Too many US politicians whine about outsourcing - they should only imagine what would have happened if India's telecoms had been liberalized five years earlier and caught more of the 1990s boom.)
The small appliances are cheap, compared to most general-purpose low power x86 servers. It's hard to build a "decent" system for under $150-200, and it's generally going to include a graphics system and want a keyboard and mouse, while you can get a typical appliance for $29-59 including a 4-port or 8-port hub. It's not as flexible, so you probably wouldn't use it for your servers, but it's a good start for protecting client-side users.
Does anybody know if you can boot Knoppix from external DVDs using USB2, assuming your BIOS is new enough to do USB booting at all? My main desktop only has a CD burner right now, and iTunes complains about not being able to read audio from it, so I'm trying to decide whether to buy a cheaper internal DVD burner or a slightly more expensive USB2 version, which would let me use the burner on my laptop which only has DVD-ROM capabilities. (I suppose I could add Firewire to the box as an alternative, but I don't think my BIOS has any chance of booting from that.)
Also, I'm curious if the files are compressed? It's a tradeoff of storage and disk read speed vs. CPU speed, and unlike CDs, I'm not sure if it's a win, but DVDs only have ~4.7GB for regular or 8.5 for double-layer, so you'll run out of space after a while :-)
Of course, so are motherboards.
It's probably too old a BIOS to boot from USB or Firewire?
- Overspecification - The customer lists *all* the specs they like from *all* the products on the market, even though no single product meets all those specs.
- Purchasing Bureaucrat requirements filtering and process formality - The vendor doesn't actually get to talk to the end user who wants the product - the vendor talks to a bunch of purchasing bureaucrats who only have a vague notion of what the end user wants, so they pad some of the requirements to be sure, and the end user has usually padded some of the requirements because he knows that the purchasing bureaucrats don't have a clue what they're doing. And the vendor can't simply sit down and talk to the end user about it, because purchasing rules usually prevent it (to prevent a whole 'nother set of ways that vendors can pad costs and rip off the government.) And usually some PHB in the process read some article in Jet Engine World or Hammer Manufacturer's Digest about some cool new innovation in jet engine hammers and wants to make sure that the end user's don't get ripped off by some vendor selling them an inferior hardware-store hammer.
- Cost Accounting for Contract Issues. When you need to buy a $5m jet engine and a $5 jet engine hammer, it's going to cost at least $1000 for the paperwork, and that gets accounted for as part of the items being purchased. So you end up with two line items in the contract, a $5,000,500 jet engine and a $505 hammer.
- Fixed vs. Variable Manufacturing Costs and Small Production Runs - Most manufacturing processes have a fixed cost to set up the manufacturing equipment and a variable cost that's the per-unit cost of making the items. So if it costs $5000 to set up the hammer-making equipment and $1 in variable costs for each hammer, Walmart is going to make a million hammers at a cost of $1+(5000/1000000) = $1.0005 each, mark them up to $10, and hold a big price-rollback sale with them marked down to $5 (which includes about a buck a hammer for advertising.) But the Air Force is going to order one mil-spec jet engine hammer with each of the ten jet engines they're buying, so they're going to pay $1 + (5000/10) = $501 per hammer, and then they're going to order a dozen mil-spec helicopter rotor hammers, which are slightly different from jet-engine hammers, so the machines need to be set up again, and they'll pay $1 + (5000/12) = $417.66 per hammer.
Would it be cheaper to order a slightly *better* hammer that can be used for both, at a cost of $2 + (5000/22) = $229/hammer? Yup, but nobody could justify the cost of doing the paperwork to explain why they're buying fancier hammers, leave aside the costs of convening a MIL-SPEC-Joint-Tactical-Hammer Design Committee, or dealing with the auditors who have to explain why Halliburton is importing thousands of $100 MIL-SPEC-Joint-Tactical-Hammer units to Iraq where they're going to use them for busting up sheetrock (which could be done with $5 Walmart hammers) or using them in place of the $500 MIL-Spec-Titanium-Tank-Wheel-Hammer which is made by a contractor in Senator Foghorn's district.The Bunker is a data center housed in a former missile base in the UK. It's underground, which makes cooling much easier, and has highly redundant commercial power feeds in addition to the stuff they've added themselves. Being able to recycle that kind of space kept their construction costs low and keeps their cooling costs low, and it also looks cool, which has been useful for them to sell service to bankers and other reliability-paranoid types.
When I moved to the Bay Area in 1993, there was one gas station in Livermore that usually had gas for prices between 0.85 to $1, and prices stayed relatively low for a few years after that in spite of California's increases in gas tax. Then *you* showed up....
Unfortunately, there are a bunch of applications for which this doesn't work right, including iTunes - the first piece of Apple software I've used that didn't "just work". When I installed iTunes, as root, it created an iTunes config for root, but when I logged in as myself, it created a separate iTunes config for me, and I not only had to input lots of long registration numbers again (:-), but the tunes I'd downloaded to root's account aren't accessible from my account and vice versa (or at least, it's well hidden if they are.) Very annoying.
Some things are worse about multiple users - my USB scanner gets hopelessly confused by having multiple people logged in. As far as I can tell, when I first log in as one user, its software scans the USB and finds it, and when I log in as a different user, it does the same thing, except something's locked up to the first person who logged in.
(As somebody else said about their home setup, I've got three accounts on the machine - root, my non-admin account, and my wife's account, which has admin privileges so she can install software and run picky software, and we use fast-user-switching between them.)
I'm not sure if that's always true, but it's certainly true for the VPN software version I use to access the network at work. Very, Very Annoying! It means that I can no longer set up my home PC to access the VPN the way I did when the home machine ran Win98/WinME, since I use XP's fast user switching between root, my non-priv account, and my wife's account.
The difference is that you're setting things up so one badly-written application can run as Admin when the user wants it, but most of your other programs are running as non-privileged-user, vs. setting the user to be Admin for everything they run. Sure, when you're running that one program, you're still exposed, but at least for most of your applications, they're less dangerous.
On the other hand, a friend of mine has the Sprint low-cost-unlimited-data service from early on that they've stopped selling (I think he's paying $30/month, maybe less) and it works just fine.
So put your OS and commonly used read-mostly files on the flash, and mount a piece of rotating machinery for /home or whatever, and either don't swap or else swap to the disk drive. If you're paranoid, build a file system that writes to disk in parallel to the flash, so you won't lose anything.
Writing flash may be slow, but reading is fast, and there's no rotational latency to worry about. So put your OS and applications in the flash, the way you would with CD for Knoppix, and that stuff will be a LOT faster. Maybe you use flash for writes, or maybe use a hard drive to mount /usr, or maybe do some sort of translucent file system thing that initially writes to disk and then migrates stuff up to flash.
Yes, writing to flash memory is generally slow. But reading is really fast, and you don't need to wait for rotating machinery to get to the right location to start, so it can really be *much* faster, and you get most of the power and performance savings by speeding up reads. If you're using USB-based flash sticks, obviously you're limited by the quality of the USB drivers, but USB2 itself is faster than most disk busses and isn't a bottleneck, though some people have argued that Firewire is lighter and faster.
Also, for performance, disk drives have rotational latency, and flash doesn't, so even if reads aren't quite as fast (especially compressed reads that use yor CPU), the effective speed is generally a lot better.
It's nice to get blazingly fast bandwidth to a handheld device, but unless it's got a hard disk bigger than 40GB and you're running BitTorrent, it's fundamentally really low usage for the wireless company. Many of the cellular companies that offer $20 "unlimited" service will not let you use that service to connect your phone/handheld to a PC, or let you use it with a cellular card in your PC, or at least won't let you do it without charging you $80-100. So that phone with a Bluetooth in it isn't allowed to use the Bluetooth to connect to your laptop..
A friend of mine lived in an area that didn't have DSL yet, but he could get the ~56kbps Metricom service, and was the only customer near his nearest Metricom lamppost, so he usually got pretty full bandwidth for his household of four people. Sure beat dialup.