A couple of years back, my work PC had fried itself, losing all my files. Oh, well, at least my calendar and contact lists were in my trust Palm Pilot. As I was driving to work a day or two later, my coffee cup bounced off the dashboard and landed in my briefcase. The Palm Pilot did not like this:-( Furthermore, my other backup set of information was the pile of Dead Trees in my briefcase, which also did not like having coffee poured on them.
So I wandered aimlessly the next couple of days, as the people that I support gradually told me what meetings they had planned for me. And Fry's had a sale on Palm Pilots.
My departmental minicomputer job in those days was an IBM System 34 at a small steel company. It had a 13MB Winchester and 48KB of semiconductor RAM (woo-who!.) The clerk had spent 6 hours typing in all the steel bars for a project, and some guy out in the shop needed to find the circuit breaker for his welder, and got ours first. The file system on those wasn't very bright - when you closed a file, it wrote down where everything was. Fortunately, the clerk had typed in an hour's worth of steel bars the day before, so it knew where the _beginning_ of the file was, and I spent about 5 hours on the phone with IBM doing the equivalent of "ed/dev/hda1" while we found all the pieces and told the machine where the end of the file was.
Back when I was a newbie, I had a file named "*" show up in my directory. So I removed it. And realized how dumb I had just been. Fortunately, this was back in the days of timesharing, so the machine actually had an administrator, and I was pleased to discover that they did nightly backups.
I actually did it on / , not.. - we had a loaner machine that we needed to return to the vendor, so I had to clean up the files anyway. Might as well have some fun with it. I was logged in on the serial port console (this _was_ the late 80s, after all.) Fairly quickly, "ls" went away, but "echo *" worked fine. The/bin directory was still there, since/bin/sh was running, but it was mostly empty. There was no/proc back then, so "ps" also went away, but I'd stopped most processes first. There was some network daemon running that kept/etc from disappearing, but most other things vanished.
YDI has been acquiring interesting companies lately. Terabeam does short-haul optical stuff, and Karl wrote the Karlbridge IP stuff a while back. They seem to be making money (???!!!) and just got new management to work with their expansions.
The blurb announces that their new stuff has proven to be stable in small trials. Cool! Outstanding! If it really is stable, and they've also done a lot of modelling work to figure out how it scales to really huge networks, then they're probably almost ready to try it out on medium-sized networks and see if their theoretical models are vaguely correct. I won't say it's easy to make things be stable in small networks, because even those can go appallingly wrong and buggy, but working on a small scale is no guarantee of not exploding in bizarre ways when tried on a larger scale.
I wasn't sure if that was some addressing thing, or if that was really some Chinese Ideogram translation thing (double-wide Big5, anybody??.) All sounded fuzzy and non-technical to me.
Most car stereos I've seen or owned had an Auxiliary Input Jack that you could use to plug in a cable from another device. They're not always installed correctly in the dashboard, but they're usually there somewhere. That's especially useful for something like an iPod that you want to take with you rather than permanently mounting in the car, and if do you want to hide it in your glove compartment, that's a good match for the aux jack that was left behind the dashboard because the stereo installers didn't feel like mounting it in the dashboard.
Another relatively simple approach, if your car stereo has a cassette player, is those adapters that look like a cassette tape with a wire out the back. Not sure how good they sound, though.
320x240 is great if you're comparing it to a PalmPilot. It's not enough to play TV on the LCD. And it's certainly not enough to run much text on, and it sounds like it can't drive a TV at good enough rates to use that as a monitor either - plus it doesn't say anything about driving HDTV so I assume it can't do that either.
BBC article, NCSA Web page. The NCSA built a supercomputer out of a cluster of 70 Playstation 2 computers. They're not actually using the main CPU for numbercrunching - they're using the Emotion Engine graphics chip.
It's normally good practice not to attribute to malice that which can be adequately explained by stupidity, but I think both are happening here (and besides, this _is_ the Ashcroft administration.)
If their database truly risks crashing, self-destruction, and Bad Things happening by allowing the public to retrieve the public's data from it, then they're being totally irresponsible by not maintaining adequate backups. If the news articles are correct and this is running in some proprietary system on a Windows 95 machine, then all they need to do is back up the disk drive to another machine (if necessary by plugging the drive into another machine.) Then make a copy of that copy and go to work - that will give them a safe environment in which to extract data from their currently write-only database environment, because if it crashes, you re-ghost the test drive and try again.
Furthermore, they can copy it to a newer machine that has a CD-ROM or DVD writer, and burn copies for the public at less than $0.50 for the entire database, as opposed to $0.50 per page of dead trees. They might argue that the data is in some proprietary database format and copying it would violate the license - but this is clearly Fair Use, if they need to make backups of the data and aren't able to do so without risking crashing the system.
If the newspaper articles are incorrect and the database is on some antique mainframe-like thing, not a Win95 box, then they still need to extract a backup copy and make copies of the backup copy available for forensic analysis. If nothing else, running "strings" on the data would be a good start...
Of course, that doesn't mean they can expect the copy to survive Slashdotting...
Apparently they're a virtual web hosting / colo company that runs on a reseller market, so it does take a fair bit of coordination about who's controlling which parts, including who's controlling what parts of DNS. So it's not simply something that NAT can fix, though their system was apparently not organized very well and it's taken them six months to change over half their infrastructure.
Now that I've read the plaintiff's documents and the TRO, it appears that they're not being as greedy as their ISP's complaint made it sound. They're asking for the IP addresses for about 60-90 days (depending on exactly which day you count their service as stopping.) It's a bit long, compared to the typical 30 days grace period that most ISPs give departing customers, but it's not a permament thing.
Not surprisingly, this case is really mostly a dispute about money.
The problem isn't just that the micreant stole a little bit of IP address space. The ISP probably has enough to get by for a while, and if not they can get more from ARIN. The problem is that the routing tables for the Internet would be absolutely huge if everybody used their own portable address space - the way they stay sane is route summarization, which is mainly ISPs advertising big chunks of space (typically a/19 or shorter prefix) and taking care of the routing details for their customers inside that without exposing them to the outside world. Because this miscreant is using a chunk of his old ISP's non-portable address space, there'll be one BGP route announcement from the ISP for their big chunk of space, and another announcement from the miscreant for their little chunk of space somewhere else, and routers around the Internet will be expected to keep track of it. (Some ISPs have a policy that they'll only accept announcements for large blocks of non-portable ISP space and small blocks of portable ARIN-issued space, and won't accept advertisements for small chunks of non-portable space, which is usually only done by people who are dual-homing for reliability.)
IPv6 was designed with the intent of supporting a variety of types of addressing and route summarization so they could avoid the rapid growth in number of routes on the net, using tricks like geographically meaningful routing, but for the most part it hasn't happened.
It's fairly common for ISPs to let ex-customers use their non-portable address space for a month or so for a transition period, but that's not what this miscreant wants - they want to keep the space permanently, because for some reason they think it's cheaper and easier to hire an aggressive lawyer than a competent sysadmin.
Your post may cost the net hundreds, if not thousands, of dollars. So go RTFA.
Nobody has to do anything to implement this - in particular, NAC, who own the address space, probably has to do precisely nothing, as opposed to recycling the space which they would normally do. (The main exception is that if they run spoof-proofing filters to reject outsiders claiming to own their address space, they'll have to make a table entry to permit this one.)
The BGP routing protocol always uses the longest available prefix unless you tell it not to, so if NAC has the IP address space x.y.*.* and the miscreant plaintiff was using x.y.z.*, the miscreant's new ISP will need to advertise that they've got a route to x.y.z/24 - and the rest of the Internet will generally accept it and send packets to new-isp.net instead of old-isp.net, running a bit slower because there's more crap cluttering up their routing tables. Some ISPs will filter out the announcement because it's too short, so they'll still send any packets for x.y.z.w to the ISP handling x.y.*.* (that's old-isp.net) who will forward it to new-isp.net.
Any equipment that's too much trouble to renumber is too old to keep using - they'd have been better off paying a network administrator for a day rather than a lawyer. They've had their own ARIN allocation for over a year and haven't used it.
There's some software out there that doesn't use DNS. Mostly it's things like VPNs that authenticate on an IP address basis because they need to accept incoming packets that don't have friendly user information with them (or couldn't trust it if they did.) I don't know if that's why this miscreant wants their IP address to stay the same, but they've had an ARIN-allocated IP address space for over a year and not renumbered.
Any exposure is bad unless you've got really heavy amounts of shielding. You get some exposure on the week-long trip up, and some on the week-long trip down, and you could avoid the latter by taking a lander instead. The asymmetry that I'd missed the first time was that you can run the elevator faster going down than going up, if you really don't like taking a lander.
Options are only worth something if the stock is expected to be worth more than the option price - otherwise they're only worth gambling money. And when they're compensation, there are usually limits on when you can exercise them - lots of dot-bomb folks had options on stock that was worth tons of money on IPO day and worthless six months later. A few years ago, my company acknowledged that their stock will probably never hit $55 again, and bought back our bonus options for something like 20:1 in stock (and something like 10:1 for the $25 options - after adjustment, the stock's worth about $4.) So those options worth "$100M" in stock - they could have been worth $100M, but they could have been really worth $10M before the merger and $0M after.
I do understand a lot about traffic shaping... The average BitTorrent user isn't behind a full-scale Cisco router. They've got some stripped-down DSL or cable modem box, possibly a separate firewall box, and one or often more computers behind it. Most consumer-DSL price hardware doesn't have much traffic-shaping capability, so if you've got two PCs fighting it out, you're going to lose. (You've got more control if you're trying to do things inside one Linux box.) Also, standard IP type of service bits let you mark stuff as more important than routine, but not as _less_ important, so your router is going to have to recognize the low-priority applications by other means, such as TCP/UDP port - that means you're going to have to configure it to know what BitTorrent looks like. (BitTorrent's not too bad for that, but there are other file-sharing protocols that deserve low priority and are harder to recognize, especially if they're deliberately trying to look like Port 80 or Port 443 or other innocuous material.)
The problem is radiation exposure, which you want to minimize - an extra week's trip means you either need to bring down _lots_ of shielding or else you're increasing your risk. On the other hand, I was assuming that the down elevator would be the same slow speed as the up elevator, which isn't necessarily so.
Actually, you would leave most of it at the top - it's much easier to get down the quick way, because a landing craft is much simpler than a rocket that has to get up into orbit. So you'd haul a bunch of landing craft up the elevator, and people who want to head back down take the express lander. Not sure if the right design is mostly a parachute or more likely mostly a glider, but it means you don't have to take a long slow trip through the Van Allen Belts - just a quick drop.
Hauling humans by elevator or rocket is a tradeoff that'll change as the thing gets deployed, but the biggest job for this thing is obviously cargo; any reasonable amount of space construction would have far more mass of stuff than bodies up there. Get an initial construction crew up and keep them there for a while, but they're going to need a lot of building material and then lots of spaceship parts.
Edwards thinks we'll have the right kind of carbon nanotube material Real Soon Now, if he gets his grant for a couple million more bucks of development money. That's dirt cheap - and if he's right, it's such a revolutionary building material for non-space use that commercial companies ought to be pounding on his doorstep to invest. Now, maybe this means that commercial companies are more realistic about his chances of quick success than he is, or maybe he's had his head in the government sand too long to realize that if he's that good he shouldn't need to wait around for government grant money and should be able to go out and raise the seed money quickly and easily.
Carbon nanotubes, if the hype is real, are a much better VC investment than most of what we're still doing in Silicon Valley. So is he real or not?
When the article title is talking about impoverishing spammers, it doesn't mean making them pay for lots of stamps - it means depriving them of their source of revenue because they can't sell enough Nigerian Herbal Fake Viagra Stock to make money.
This system isn't making senders of email pay cash. It's making them burn CPU time, which isn't a problem if you're a real human sending out mail at the rates that real humans do, but it's a speed limit on the rate that spammers can send mail (e.g. 10 seconds per message means you can only send 8640 spams/day instead of 8 million.) This is a bit of an annoyance, but unless you're running a mailing list, it's not a big problem, and there are separate methods for handling mailing lists (if you want to subscribe to a list, you need to whitelist the list.)
Slashdot Mirror
So I wandered aimlessly the next couple of days, as the people that I support gradually told me what meetings they had planned for me. And Fry's had a sale on Palm Pilots.
My departmental minicomputer job in those days was an IBM System 34 at a small steel company. It had a 13MB Winchester and 48KB of semiconductor RAM (woo-who!.) The clerk had spent 6 hours typing in all the steel bars for a project, and some guy out in the shop needed to find the circuit breaker for his welder, and got ours first. The file system on those wasn't very bright - when you closed a file, it wrote down where everything was. Fortunately, the clerk had typed in an hour's worth of steel bars the day before, so it knew where the _beginning_ of the file was, and I spent about 5 hours on the phone with IBM doing the equivalent of "ed /dev/hda1" while we found all the pieces and told the machine where the end of the file was.
Back when I was a newbie, I had a file named "*" show up in my directory. So I removed it. And realized how dumb I had just been. Fortunately, this was back in the days of timesharing, so the machine actually had an administrator, and I was pleased to discover that they did nightly backups.
I actually did it on / , not .. - we had a loaner machine that we needed to return to the vendor, so I had to clean up the files anyway. Might as well have some fun with it. I was logged in on the serial port console (this _was_ the late 80s, after all.) Fairly quickly, "ls" went away, but "echo *" worked fine. The /bin directory was still there, since /bin/sh was running, but it was mostly empty. There was no /proc back then, so "ps" also went away, but I'd stopped most processes first. There was some network daemon running that kept /etc from disappearing, but most other things vanished.
YDI has been acquiring interesting companies lately. Terabeam does short-haul optical stuff, and Karl wrote the Karlbridge IP stuff a while back. They seem to be making money (???!!!) and just got new management to work with their expansions.
The blurb announces that their new stuff has proven to be stable in small trials. Cool! Outstanding! If it really is stable, and they've also done a lot of modelling work to figure out how it scales to really huge networks, then they're probably almost ready to try it out on medium-sized networks and see if their theoretical models are vaguely correct. I won't say it's easy to make things be stable in small networks, because even those can go appallingly wrong and buggy, but working on a small scale is no guarantee of not exploding in bizarre ways when tried on a larger scale.
I wasn't sure if that was some addressing thing, or if that was really some Chinese Ideogram translation thing (double-wide Big5, anybody??.) All sounded fuzzy and non-technical to me.
Another relatively simple approach, if your car stereo has a cassette player, is those adapters that look like a cassette tape with a wire out the back. Not sure how good they sound, though.
320x240 is great if you're comparing it to a PalmPilot. It's not enough to play TV on the LCD. And it's certainly not enough to run much text on, and it sounds like it can't drive a TV at good enough rates to use that as a monitor either - plus it doesn't say anything about driving HDTV so I assume it can't do that either.
BBC article, NCSA Web page. The NCSA built a supercomputer out of a cluster of 70 Playstation 2 computers. They're not actually using the main CPU for numbercrunching - they're using the Emotion Engine graphics chip.
If their database truly risks crashing, self-destruction, and Bad Things happening by allowing the public to retrieve the public's data from it, then they're being totally irresponsible by not maintaining adequate backups. If the news articles are correct and this is running in some proprietary system on a Windows 95 machine, then all they need to do is back up the disk drive to another machine (if necessary by plugging the drive into another machine.) Then make a copy of that copy and go to work - that will give them a safe environment in which to extract data from their currently write-only database environment, because if it crashes, you re-ghost the test drive and try again.
Furthermore, they can copy it to a newer machine that has a CD-ROM or DVD writer, and burn copies for the public at less than $0.50 for the entire database, as opposed to $0.50 per page of dead trees.
They might argue that the data is in some proprietary database format and copying it would violate the license - but this is clearly Fair Use, if they need to make backups of the data and aren't able to do so without risking crashing the system.
If the newspaper articles are incorrect and the database is on some antique mainframe-like thing, not a Win95 box, then they still need to extract a backup copy and make copies of the backup copy available for forensic analysis. If nothing else, running "strings" on the data would be a good start...
Of course, that doesn't mean they can expect the copy to survive Slashdotting...
Apparently they're a virtual web hosting / colo company that runs on a reseller market, so it does take a fair bit of coordination about who's controlling which parts, including who's controlling what parts of DNS. So it's not simply something that NAT can fix, though their system was apparently not organized very well and it's taken them six months to change over half their infrastructure.
Not surprisingly, this case is really mostly a dispute about money.
IPv6 was designed with the intent of supporting a variety of types of addressing and route summarization so they could avoid the rapid growth in number of routes on the net, using tricks like geographically meaningful routing, but for the most part it hasn't happened.
It's fairly common for ISPs to let ex-customers use their non-portable address space for a month or so for a transition period, but that's not what this miscreant wants - they want to keep the space permanently, because for some reason they think it's cheaper and easier to hire an aggressive lawyer than a competent sysadmin.
Nobody has to do anything to implement this - in particular, NAC, who own the address space, probably has to do precisely nothing, as opposed to recycling the space which they would normally do. (The main exception is that if they run spoof-proofing filters to reject outsiders claiming to own their address space, they'll have to make a table entry to permit this one.)
The BGP routing protocol always uses the longest available prefix unless you tell it not to, so if NAC has the IP address space x.y.*.* and the miscreant plaintiff was using x.y.z.*, the miscreant's new ISP will need to advertise that they've got a route to x.y.z/24 - and the rest of the Internet will generally accept it and send packets to new-isp.net instead of old-isp.net, running a bit slower because there's more crap cluttering up their routing tables. Some ISPs will filter out the announcement because it's too short, so they'll still send any packets for x.y.z.w to the ISP handling x.y.*.* (that's old-isp.net) who will forward it to new-isp.net.
Any equipment that's too much trouble to renumber is too old to keep using - they'd have been better off paying a network administrator for a day rather than a lawyer. They've had their own ARIN allocation for over a year and haven't used it.
There's some software out there that doesn't use DNS. Mostly it's things like VPNs that authenticate on an IP address basis because they need to accept incoming packets that don't have friendly user information with them (or couldn't trust it if they did.) I don't know if that's why this miscreant wants their IP address to stay the same, but they've had an ARIN-allocated IP address space for over a year and not renumbered.
Any exposure is bad unless you've got really heavy amounts of shielding. You get some exposure on the week-long trip up, and some on the week-long trip down, and you could avoid the latter by taking a lander instead. The asymmetry that I'd missed the first time was that you can run the elevator faster going down than going up, if you really don't like taking a lander.
Besides, some people _like_ ranting :-)
I do understand a lot about traffic shaping... The average BitTorrent user isn't behind a full-scale Cisco router. They've got some stripped-down DSL or cable modem box, possibly a separate firewall box, and one or often more computers behind it. Most consumer-DSL price hardware doesn't have much traffic-shaping capability, so if you've got two PCs fighting it out, you're going to lose. (You've got more control if you're trying to do things inside one Linux box.) Also, standard IP type of service bits let you mark stuff as more important than routine, but not as _less_ important, so your router is going to have to recognize the low-priority applications by other means, such as TCP/UDP port - that means you're going to have to configure it to know what BitTorrent looks like. (BitTorrent's not too bad for that, but there are other file-sharing protocols that deserve low priority and are harder to recognize, especially if they're deliberately trying to look like Port 80 or Port 443 or other innocuous material.)
The problem is radiation exposure, which you want to minimize - an extra week's trip means you either need to bring down _lots_ of shielding or else you're increasing your risk. On the other hand, I was assuming that the down elevator would be the same slow speed as the up elevator, which isn't necessarily so.
Actually, you would leave most of it at the top - it's much easier to get down the quick way, because a landing craft is much simpler than a rocket that has to get up into orbit. So you'd haul a bunch of landing craft up the elevator, and people who want to head back down take the express lander. Not sure if the right design is mostly a parachute or more likely mostly a glider, but it means you don't have to take a long slow trip through the Van Allen Belts - just a quick drop.
Hauling humans by elevator or rocket is a tradeoff that'll change as the thing gets deployed, but the biggest job for this thing is obviously cargo; any reasonable amount of space construction would have far more mass of stuff than bodies up there. Get an initial construction crew up and keep them there for a while, but they're going to need a lot of building material and then lots of spaceship parts.
Carbon nanotubes, if the hype is real, are a much better VC investment than most of what we're still doing in Silicon Valley. So is he real or not?
This system isn't making senders of email pay cash. It's making them burn CPU time, which isn't a problem if you're a real human sending out mail at the rates that real humans do, but it's a speed limit on the rate that spammers can send mail (e.g. 10 seconds per message means you can only send 8640 spams/day instead of 8 million.) This is a bit of an annoyance, but unless you're running a mailing list, it's not a big problem, and there are separate methods for handling mailing lists (if you want to subscribe to a list, you need to whitelist the list.)