No, but they strongly recommend it. Words like "obsolete" are being bandied around, and I've known people not help in fora unless you're using devfs.
Well, if you need to be looking for help in forums you should probably not worry this issue at all. You should just use what your distro uses. If your distro wants to continue to use devfs they can attempt to patch it into future kernels. It is something you shouldn't have to worry about.
That's not the case with XP. I've tested it out -- use the install disc to format to NTFS, then switch over to Linux. It'll ID it correctly as NTFS.
Where exactly does it say that the partition ID has to match the filesystem that is currently on the drive? Did you try actually mounting it to make sure?
I have no idea what Windows does at that point in the install process, and it really make no difference at all.
I'm not, I'm opposed to having to switch around my OS every other week. And I saw no benefit of devfs or udev over a simple static/dev.
Then don't change it around. Nobody forced you to use devfs. Nobody is forcing you to use udev. If you like the old school way of doing things, don't change.
This is flawed. You can't just divide there. You are estimating he gets 5 passengers per 1/10 mach, which would be 0.5 passengers per mach.
If you use your math, but extend the capacity of his car up to 200 (just like this Japanese plane) you will get 2000 passengers/mach. (200 / 0.1 == 2000)
How about mach 1 == 765.6 mph. (65/765.6) * 5 ==.42450365726227795190 passengers/mach
Be glad you live on that side of the pond. If we continental europeans switch our PSUs from 220V to 110V, we get a nice, smoke-filled room with a side dish of fried motherboard.
Funny, that is exactly what I surmised would have happened that very day:). Although it would have saved me a whole lot of trouble if we did live in Europe... The thing would have been fried before I got over there:p.
I do not know if this is actually a mishap or not, but it is one of my favorite stories. Sometime about 10 years or so ago, during high school, a friend of mine was building a computer. I do not actually recall if it was for himself or not, but I believe it was a 486 25 or 33 mhz or so.
He just couldn't get it to work at all, and asked if I could stop by and help him out. When I got there, the machine would power up, and the power supply fan was spinning just fine.
I recall I started with easy things like reseating the memory, reseating ISA cards... When none of that worked, I disassembled the whole thing and put it back together. Same symptoms as before. He tried similar things, same problems.
I was sitting staring at the machine... And I saw the problem. I told him I knew exactly what was wrong, but I told him I shouldn't tell him, and I should let him find it himself.
I did end up telling him... The power supply voltage was set to 220 instead of 110...
I'm confused... you want the OP to use striped disks, yet you want the OP to have data integrity as a primary goal. That's like asking for a raw boiled egg.
Striping with data integrity is called RAID 10 or RAID 5, both of which provide better data integrity than a single disk. RAID 10 if you need good write performance, and RAID 5 if you are on a budget and only need good read performance. Heck, even mirroring will increase read speed.
You obviously have never done contract work in the field. The ability to ping people you are working with real quick without making an official phone call is great.
Text messaging is usually good for such things. The receiver does not have to respond right away, and they can read the message more than once. I would imagine a vibrating phone/2-way is less of a disturbance than a phone going '*BLEEP BLEEP* Are you there?'
You may even be able to send out a quick text reply without interrupting anyone else.
You don't ping your buddy at the restaurant to see if he wants to go to the bar tonight.
Actually, text messaging is quite good for this. I do not need to know that he is going right this minute. He can likely tell me any time in the next few hours. Of course, I have no idea if he is in a meeting, napping, out to lunch with someone important. I do not need to know, he can respond any time.
Please correct me if I am mistaken, but it seems to me that text messaging and push to talk are both best used for asyncronous communications. Don't push to talk messages disappear immediately after you hear them, or are they saved? I am under the impression that they disappear. For me, that would severely limit its uses...
I have to admit though, text messaging was much easier on my old Motorola 2 way pager than it is on the average cell phone. I can probably thumb about 4-6 characters per second on those things... I imagine I am more like 1-2 on a phone:).
I would hope it would die, but unfortunately Sprint happens to have the same feature already anyway...
Why do people bother with this? Why don't they just call each other? It is bad enough when I have to listen to half of a conversation. Now I get to sit in a restaraunt, or other public place, and hear:
Phone: *BLEEP BLEEP* Hey, are you there?
Guy: Yeah, I am here
Phone: *BLEEP BLEEP* Where are you?
Guy: I am at a restaurant.
Phone: *BLEEP BLEEP* What did you say?
Guy: AT A RESTAURANT
Phone: *BLEEP BLEEP* Oh! At a restaurant!
Does Nextel not give free mobile to mobile minutes? Do people not notice that they make their conversations so public like this?
Lots of Nextel customers I have met actually think their phone works like a walkie talkie. They are amazed that it goes 300 miles, directly to the other cell!
No, USB is a completely different and far more difficult issue to handle.
It is not really COMPLETELY different... USB may have other uses, but on a corporate desktop you are only likely to use USB for keyboards and mice.
With floppies, tapes, CD-ROMs etc, it is easy to restrict a PC. The peripherals can either be removed completely or they can have physical locks placed on them that require a key in order to use them. The peripherals can also be disabled in the BIOS which in turn can be protected by password. So, with these devices, it is relatively easy to prevent users from using them at all.
This is all technically true of USB as well. I have never specifically looked, but there is no technical reason you cannot manufacture a physical lock for a USB port. Lock in a keyboard and mouse (or use PS/2 instead) and you are set. And as you said, you can disable USB in the BIOS.
If your company is this concerned about data security then they should buy machines with PS/2 keyboards and mice.
However, most companies have still bought machines with floppy drives for the past 20 years. If they were not worried about this problem then, why would they be worried today?
So, the problem is a massive one. How do you limit the connection of certain USB devices, such as flash drives or WiFi dongles, to the machines on your network while still allowing most other devices to function?
If this is the issue, why not just remove all USB drivers from the system except for HID devices? I would imagine a USB drive would not work without the mass storage driver installed.
The problem may be as small as a 512MB keychain fob or as large as a 300GB external USB hard drive hidden in a purse.
Go back 10 years and substitute "keychain fob" with "floppy diskette." We have had this problem forever, and it is not new. If a company was truly concerned about this they would buy machines with no removable storage that was writeable. They can do the same today by going PS/2 for keyboards and mice. These companies will not care if they have to pay extra.
Connecting a USB WiFi fob in a multi-story building is another monsterous security issue.
Again, removing the drivers will fix this problem. You still, however, need to worry about someone plugging wireless bridges into your network... I can drop a hub and a wireless bridge under your secretary's desk. Then all I'd have to do is spoof her MAC address at night and poke around your network all I want. You would likely never notice it was happening until it was too late.
In any case, USB security is different than floppies and CD-RWs and it is a serious matter for those that are concerned with security.
It is a SLIGHTLY different problem with very similar solutions.
src/router/www - html, css, javascript, images for web control
Hmmm... All the html I have seen on the Sveasoft firmware looks like a derivitive of the Linksys html. I have absolutely no idea what license Linksys actually release those files under. Was it a license that allows modification and redistribution?
Aside from that, they still need to provide source for the GPL and LGPL components. If they do not they are distributing it illegally. It doesn't matter that some of the puzzle pieces are legal or not.
My bandwidth tests with comcast always came back 5.1 MB down/28-37 KB up.
You just answered your own question. You're most definitely not getting 5 megaBYTES downstream (that would be 40 megabit). I will assume you also mixed up your upstream bits vs. bytes. 28-37KB would be 224 - 296 kilobits. I believe Comcast bumped everyone to either 256 or 384 kilobit upstream.
My virtual host? I have had it for about a year now. I believe it has been down twice for maintenence. It currently has 114 days uptime. I also do not have to worry about backups, they do it for me. On the other hand my cable modem has been down/unresponsive for more short periods than I care to count. They do not, and will never guaranty 5 nines of uptime. My virtual host does. If I want that out of Cable/DSL I would need to pay more.
Or if you wish to update, what then?
With the server at your house, YOU are in control. Not the virtual hosting.
My server came with Debian Woody, I updated it to Sarge. I have root, I can do most anything I want. If I need more memory, disk, or bandwidth I just buy more.
As it is, the DSL and Cable industry are getting ready to do another bump up (to support VOIP). At that point, then you will have loads of speed to your system. As long as it is used for personal use, most will not care.
First of all, I do not have that speed now. Second, I do not expect the same uptime out of a residential service.
I host domains for myself and a few friends. None of them, especially myself, would be happy to have an email outage:p.
39.99+all sorts of fees = 54.00 / mo cable does NOT allow me to run ANY servers, and block most of the default service ports for unix... (most still allow windows, but I'm not about to buy IIS to run a simple site on that huge clunking POS).
In most areas DSL is much slower than cable. My cable here is somewhere in the neighborhood of 3 megabit down and 128 up. When I had Comcast they bumped it up to something like 4/256 or some such.
Why on earth would you want to run any services over DSL or cable? The speed and reliabilty are absolutely awful. I have a little Debian virtual server that I pay somewhere short of 20 dollars a month for. I get 30 gigabytes of bandwidth per month and I have seen sustained uplink speeds of over 500 KB/second.
I certainly do not have the cheapest package with my provider, and there are certainly cheaper providers.
It was the 8086 that was designed in Israel. The 8086 was a cheaper version of the 8088. For instance, it used an 8-bit data bus rather than a 16-bit bus (but internally it was the same). It was chosen for the IBM PC, due to these cost advantages.
I believe you meant to say that the 8088 was the cheaper version of the 8086.
The 8088 had the 8 bit bus. 8088 is to 80886 as 386sx is to 386dx:).
If we're talking about a cache of data we've already read incase we need it again, that's one thing, but I'm talking about a read-ahead cache.
I assumed that is what we were talking about. Either way you are likely to get an improvement in performance, though. The caching mechanism in your OS should not attempt to read ahead while another read is pending, so you aren't wasting time.
A read-ahead cache, as I understand it, normally tries to anticipate what an application might want, and reads it ahead of time.
A read ahead cache generally reads consecutive sectors following what it is currently reading. Since the head is already in place and ready for this, it is a good idea. In this case it would be much, much less helpful. On this particular implementation I doubt it could hurt because this device 'only' seems to have a seek time about 3 times faster than the Raptor (I am just making an educated guess based on the available benchmarks).
Caching will still be a huge help on both reads and writes, though. PC100 has a throughput of 800MB/sec and it just gets better from there. You can assume that any machine you are likely to put this in will have much more memory bandwidth than that.
I still believe the biggest problem with this device is that is still has a seek time that can likely be measured in milliseconds. The big advantage of a memory based drive is supposed to be seek times measured in nanoseconds. Defragging should be virtually useless on a memory based drive.
I would be very curious how the seek time on this would compare to the Raptor they tested if you only partitioned the first 4 GB. That would keep the heads always within a very short distance from the next seek, and your only real latency on a seek would be the revolution of the drive.
I would imagine that this drive would be faster with caches off.
You are probably incorrect. System memory will have much lower latency and higher throughput than SATA. Even PC100 would be quite a few times faster than the 100MB/sec peak performance they got out of this device.
they tell nothing about what the impact will be of dropping seek time from several milliseconds to, essentially, zero.
They don't tell you about dropping the seek time to essentially zero because this thing doesn't do that. Their 300 meg Mozilla source copy should be seek-intensive and it runs about 3 times faster than the hard drive (25 seconds). The 693 meg single file copy takes 7 seconds. If we guess and say a 300 meg single file would take about 3 seconds, that means it is spending 21 seconds seeking on the Mozilla copy test.
It is a pretty darn good improvement over a standard hard drive, but still not up to ramdrive seek times.
This is not the case. You are bottlenecked by the SATA real-world throughput of 100MB/sec. I already get about 45-50MB/sec out of my 2-year-old IDE drives.
You may get that kind of throughput on sequential reads or writes, but you won't get anywhere near that on random access. The article tested copying 300 megabytes of Mozilla source code around, which is a pretty good real world test of random access speeds. Each time the copy hits a new file the drive has to seek, which takes a few milliseconds. That adds up very fast on thousands of files.
After reading the article I am quite disappointed in the random access speeds for this device. I would have hoped that random reads/writes would have been almost the same speed as sequential (copying a single large file, as an example). I assume it must be SATA overhead of some sort or their FPGA is a POS. I would be the first one to order one of these things if that improved their seek time.
Your swap partition won't do well, because adding the memory to your system will likely negate the need for a swap partition anyhow (assuming you use a sane operating system that doesn't come from anywhere near Redmond).
I doubt I would personally use this thing for a swap file, but... If you have your shiney new system that can run DDR2 1000 memory (yes, I know we don't have those yet), you can spend 100 bucks or so and use the 1-4 gig of outdated DDR you already have laying around. It would probably be worth 100 bucks plus unused parts to move swap and journals off of the old spinning disks.
Re:So use two controllers
on
Basics of RAID
·
· Score: 1
With todays feature sizes you'll be lucky to get 10 years of 24/7/365 use out of any reasonably complex integrated circuit...
Americans, especially, are totally obnoxious when it comes to languages: many of them refuse to speak any other language unless they REALLY have to
That is a completely FALSE statement. I am an American, and I absolutely REFUSE to put myself in a situation where I would be required to speak any other language.
Well, it's not so doom if you are carrying your own software around, as well - to compromise, they'd need to actively scan a program that wasn't on the machine before for a decrypted private key.
I suppose it depends on your level of paranoia and how important the machine is to you. The terminal you use doesn't need anything nearly that complex. It just needs to have a keylogger to catch your passphrase, and it needs to save any files you access on your flash drive.
Also, it strikes me that it wouldn't be horrendous to construct yourself a script that would create a new keypair for you, encrypt it with a password that you only use for those (because the computer would need to have it stored), and put the private on the thumb drive and the public into authorized_keys, replacing the old one. Have it run every time that hotplug detects that particular thumb drive, and you're good. A new keypair to use from untrusted computers within by just having your thumb drive plugged in for a bit. If I didn't have a laptop to keep my keys, programs, etc. on, I'd probably go ahead and implement that now...
You'd possibly use OPIE instead, like I suggested to Basje. I am not using it myself, so there may be security implications I am currently unaware of with it. There is an OPIE password generator for Palm OS, Windows, and Unix, so you can store those on your flash drive. It is safe to use them since the password is only good once.
Or you could try the low-tech route and write a bunch of passwords down and keep changing your password in your bash_logout:). I'd rather give Opie a shot first though, since it seemed to be stable and secure already.
Is there a way to implement one time passwords with ssh?
There is an OTP package called OPIE. I toyed with it a year or so ago, it seemed to work OK. I have not used it in production, so be sure to research it first.
There is a password generator for Palm OS called pilOTP. I played around with both OPIE and pilOTP and they seemed to work quite well.
I wouldn't mind carrying around a slip of paper with 50 or so generated passwords, to login to my server. I'd cross off one at a time, each time only allowing the next one to be able to login some maintenance account.
Neat idea, but I don't think it will work with OPIE. OPIE is a challenge/response system I believe. If you do find a system that will let you do it with paper, be sure to obfuscate it somehow. If you lose the paper you want to have time to revoke those passwords.
I do not know how secure it would be but you could probably rig up your idea with a list of passwords and some trickery in your bash_logout. I do not know if I would trust it, but it might be a fun project:)
Slashdot Mirror
No, but they strongly recommend it. Words like "obsolete" are being bandied around, and I've known people not help in fora unless you're using devfs.
Well, if you need to be looking for help in forums you should probably not worry this issue at all. You should just use what your distro uses. If your distro wants to continue to use devfs they can attempt to patch it into future kernels. It is something you shouldn't have to worry about.
That's not the case with XP. I've tested it out -- use the install disc to format to NTFS, then switch over to Linux. It'll ID it correctly as NTFS.
Where exactly does it say that the partition ID has to match the filesystem that is currently on the drive? Did you try actually mounting it to make sure?
I have no idea what Windows does at that point in the install process, and it really make no difference at all.
I'm not, I'm opposed to having to switch around my OS every other week. And I saw no benefit of devfs or udev over a simple static /dev.
Then don't change it around. Nobody forced you to use devfs. Nobody is forcing you to use udev. If you like the old school way of doing things, don't change.
5 passengers at Mach 0.1 == 5 / 0.1 == 50
This is flawed. You can't just divide there. You are estimating he gets 5 passengers per 1/10 mach, which would be 0.5 passengers per mach.
If you use your math, but extend the capacity of his car up to 200 (just like this Japanese plane) you will get 2000 passengers/mach. (200 / 0.1 == 2000)
How about mach 1 == 765.6 mph. (65/765.6) * 5 == .42450365726227795190 passengers/mach
Be glad you live on that side of the pond. If we continental europeans switch our PSUs from 220V to 110V, we get a nice, smoke-filled room with a side dish of fried motherboard.
Funny, that is exactly what I surmised would have happened that very day :). Although it would have saved me a whole lot of trouble if we did live in Europe... The thing would have been fried before I got over there :p.
That's nothing - my Toyota Camry with 5 passengers going 65mph gets 58 passengers/mach!
If you are trying to go by his metric, your Camry gets 0.42 passengers per mach at 65 mph.
I do not know if this is actually a mishap or not, but it is one of my favorite stories. Sometime about 10 years or so ago, during high school, a friend of mine was building a computer. I do not actually recall if it was for himself or not, but I believe it was a 486 25 or 33 mhz or so.
He just couldn't get it to work at all, and asked if I could stop by and help him out. When I got there, the machine would power up, and the power supply fan was spinning just fine.
I recall I started with easy things like reseating the memory, reseating ISA cards... When none of that worked, I disassembled the whole thing and put it back together. Same symptoms as before. He tried similar things, same problems.
I was sitting staring at the machine... And I saw the problem. I told him I knew exactly what was wrong, but I told him I shouldn't tell him, and I should let him find it himself.
I did end up telling him... The power supply voltage was set to 220 instead of 110...
I'm confused... you want the OP to use striped disks, yet you want the OP to have data integrity as a primary goal. That's like asking for a raw boiled egg.
Striping with data integrity is called RAID 10 or RAID 5, both of which provide better data integrity than a single disk. RAID 10 if you need good write performance, and RAID 5 if you are on a budget and only need good read performance. Heck, even mirroring will increase read speed.
Very different than a raw boiled egg...
You obviously have never done contract work in the field. The ability to ping people you are working with real quick without making an official phone call is great.
Text messaging is usually good for such things. The receiver does not have to respond right away, and they can read the message more than once. I would imagine a vibrating phone/2-way is less of a disturbance than a phone going '*BLEEP BLEEP* Are you there?'
You may even be able to send out a quick text reply without interrupting anyone else.
You don't ping your buddy at the restaurant to see if he wants to go to the bar tonight.
Actually, text messaging is quite good for this. I do not need to know that he is going right this minute. He can likely tell me any time in the next few hours. Of course, I have no idea if he is in a meeting, napping, out to lunch with someone important. I do not need to know, he can respond any time.
Please correct me if I am mistaken, but it seems to me that text messaging and push to talk are both best used for asyncronous communications. Don't push to talk messages disappear immediately after you hear them, or are they saved? I am under the impression that they disappear. For me, that would severely limit its uses...
I have to admit though, text messaging was much easier on my old Motorola 2 way pager than it is on the average cell phone. I can probably thumb about 4-6 characters per second on those things... I imagine I am more like 1-2 on a phone :).
Then what happens to Push to Talk?
I would hope it would die, but unfortunately Sprint happens to have the same feature already anyway...
Why do people bother with this? Why don't they just call each other? It is bad enough when I have to listen to half of a conversation. Now I get to sit in a restaraunt, or other public place, and hear:
Phone: *BLEEP BLEEP* Hey, are you there?
Guy: Yeah, I am here
Phone: *BLEEP BLEEP* Where are you?
Guy: I am at a restaurant.
Phone: *BLEEP BLEEP* What did you say?
Guy: AT A RESTAURANT
Phone: *BLEEP BLEEP* Oh! At a restaurant!
Does Nextel not give free mobile to mobile minutes? Do people not notice that they make their conversations so public like this?
Lots of Nextel customers I have met actually think their phone works like a walkie talkie. They are amazed that it goes 300 miles, directly to the other cell!
No, USB is a completely different and far more difficult issue to handle.
It is not really COMPLETELY different... USB may have other uses, but on a corporate desktop you are only likely to use USB for keyboards and mice.
With floppies, tapes, CD-ROMs etc, it is easy to restrict a PC. The peripherals can either be removed completely or they can have physical locks placed on them that require a key in order to use them. The peripherals can also be disabled in the BIOS which in turn can be protected by password. So, with these devices, it is relatively easy to prevent users from using them at all.
This is all technically true of USB as well. I have never specifically looked, but there is no technical reason you cannot manufacture a physical lock for a USB port. Lock in a keyboard and mouse (or use PS/2 instead) and you are set. And as you said, you can disable USB in the BIOS.
If your company is this concerned about data security then they should buy machines with PS/2 keyboards and mice.
However, most companies have still bought machines with floppy drives for the past 20 years. If they were not worried about this problem then, why would they be worried today?
So, the problem is a massive one. How do you limit the connection of certain USB devices, such as flash drives or WiFi dongles, to the machines on your network while still allowing most other devices to function?
If this is the issue, why not just remove all USB drivers from the system except for HID devices? I would imagine a USB drive would not work without the mass storage driver installed.
The problem may be as small as a 512MB keychain fob or as large as a 300GB external USB hard drive hidden in a purse.
Go back 10 years and substitute "keychain fob" with "floppy diskette." We have had this problem forever, and it is not new. If a company was truly concerned about this they would buy machines with no removable storage that was writeable. They can do the same today by going PS/2 for keyboards and mice. These companies will not care if they have to pay extra.
Connecting a USB WiFi fob in a multi-story building is another monsterous security issue.
Again, removing the drivers will fix this problem. You still, however, need to worry about someone plugging wireless bridges into your network... I can drop a hub and a wireless bridge under your secretary's desk. Then all I'd have to do is spoof her MAC address at night and poke around your network all I want. You would likely never notice it was happening until it was too late.
In any case, USB security is different than floppies and CD-RWs and it is a serious matter for those that are concerned with security.
It is a SLIGHTLY different problem with very similar solutions.
src/router/www - html, css, javascript, images for web control
Hmmm... All the html I have seen on the Sveasoft firmware looks like a derivitive of the Linksys html. I have absolutely no idea what license Linksys actually release those files under. Was it a license that allows modification and redistribution?
Aside from that, they still need to provide source for the GPL and LGPL components. If they do not they are distributing it illegally. It doesn't matter that some of the puzzle pieces are legal or not.
My bandwidth tests with comcast always came back 5.1 MB down/28-37 KB up.
You just answered your own question. You're most definitely not getting 5 megaBYTES downstream (that would be 40 megabit). I will assume you also mixed up your upstream bits vs. bytes. 28-37KB would be 224 - 296 kilobits. I believe Comcast bumped everyone to either 256 or 384 kilobit upstream.
And when it goes down, what then?
My virtual host? I have had it for about a year now. I believe it has been down twice for maintenence. It currently has 114 days uptime. I also do not have to worry about backups, they do it for me. On the other hand my cable modem has been down/unresponsive for more short periods than I care to count. They do not, and will never guaranty 5 nines of uptime. My virtual host does. If I want that out of Cable/DSL I would need to pay more.
Or if you wish to update, what then? With the server at your house, YOU are in control. Not the virtual hosting.
My server came with Debian Woody, I updated it to Sarge. I have root, I can do most anything I want. If I need more memory, disk, or bandwidth I just buy more.
As it is, the DSL and Cable industry are getting ready to do another bump up (to support VOIP). At that point, then you will have loads of speed to your system. As long as it is used for personal use, most will not care.
First of all, I do not have that speed now. Second, I do not expect the same uptime out of a residential service.
I host domains for myself and a few friends. None of them, especially myself, would be happy to have an email outage :p.
39.99+all sorts of fees = 54.00 / mo cable does NOT allow me to run ANY servers, and block most of the default service ports for unix... (most still allow windows, but I'm not about to buy IIS to run a simple site on that huge clunking POS).
In most areas DSL is much slower than cable. My cable here is somewhere in the neighborhood of 3 megabit down and 128 up. When I had Comcast they bumped it up to something like 4/256 or some such.
Why on earth would you want to run any services over DSL or cable? The speed and reliabilty are absolutely awful. I have a little Debian virtual server that I pay somewhere short of 20 dollars a month for. I get 30 gigabytes of bandwidth per month and I have seen sustained uplink speeds of over 500 KB/second.
I certainly do not have the cheapest package with my provider, and there are certainly cheaper providers.
And I believe you meant to say "8088 is to 8086"
I can't believe I missed that! Even with brand new glasses :p.
but then, who actually checks their posts these days? :-)
Nobody is perfect :).
It was the 8086 that was designed in Israel. The 8086 was a cheaper version of the 8088. For instance, it used an 8-bit data bus rather than a 16-bit bus (but internally it was the same). It was chosen for the IBM PC, due to these cost advantages.
I believe you meant to say that the 8088 was the cheaper version of the 8086.
The 8088 had the 8 bit bus. 8088 is to 80886 as 386sx is to 386dx :).
If we're talking about a cache of data we've already read incase we need it again, that's one thing, but I'm talking about a read-ahead cache.
I assumed that is what we were talking about. Either way you are likely to get an improvement in performance, though. The caching mechanism in your OS should not attempt to read ahead while another read is pending, so you aren't wasting time.
A read-ahead cache, as I understand it, normally tries to anticipate what an application might want, and reads it ahead of time.
A read ahead cache generally reads consecutive sectors following what it is currently reading. Since the head is already in place and ready for this, it is a good idea. In this case it would be much, much less helpful. On this particular implementation I doubt it could hurt because this device 'only' seems to have a seek time about 3 times faster than the Raptor (I am just making an educated guess based on the available benchmarks).
Caching will still be a huge help on both reads and writes, though. PC100 has a throughput of 800MB/sec and it just gets better from there. You can assume that any machine you are likely to put this in will have much more memory bandwidth than that.
I still believe the biggest problem with this device is that is still has a seek time that can likely be measured in milliseconds. The big advantage of a memory based drive is supposed to be seek times measured in nanoseconds. Defragging should be virtually useless on a memory based drive.
I would be very curious how the seek time on this would compare to the Raptor they tested if you only partitioned the first 4 GB. That would keep the heads always within a very short distance from the next seek, and your only real latency on a seek would be the revolution of the drive.
I would imagine that this drive would be faster with caches off.
You are probably incorrect. System memory will have much lower latency and higher throughput than SATA. Even PC100 would be quite a few times faster than the 100MB/sec peak performance they got out of this device.
they tell nothing about what the impact will be of dropping seek time from several milliseconds to, essentially, zero.
They don't tell you about dropping the seek time to essentially zero because this thing doesn't do that. Their 300 meg Mozilla source copy should be seek-intensive and it runs about 3 times faster than the hard drive (25 seconds). The 693 meg single file copy takes 7 seconds. If we guess and say a 300 meg single file would take about 3 seconds, that means it is spending 21 seconds seeking on the Mozilla copy test.
It is a pretty darn good improvement over a standard hard drive, but still not up to ramdrive seek times.
This is not the case. You are bottlenecked by the SATA real-world throughput of 100MB/sec. I already get about 45-50MB/sec out of my 2-year-old IDE drives.
You may get that kind of throughput on sequential reads or writes, but you won't get anywhere near that on random access. The article tested copying 300 megabytes of Mozilla source code around, which is a pretty good real world test of random access speeds. Each time the copy hits a new file the drive has to seek, which takes a few milliseconds. That adds up very fast on thousands of files.
After reading the article I am quite disappointed in the random access speeds for this device. I would have hoped that random reads/writes would have been almost the same speed as sequential (copying a single large file, as an example). I assume it must be SATA overhead of some sort or their FPGA is a POS. I would be the first one to order one of these things if that improved their seek time.
Your swap partition won't do well, because adding the memory to your system will likely negate the need for a swap partition anyhow (assuming you use a sane operating system that doesn't come from anywhere near Redmond).
I doubt I would personally use this thing for a swap file, but... If you have your shiney new system that can run DDR2 1000 memory (yes, I know we don't have those yet), you can spend 100 bucks or so and use the 1-4 gig of outdated DDR you already have laying around. It would probably be worth 100 bucks plus unused parts to move swap and journals off of the old spinning disks.
With todays feature sizes you'll be lucky to get 10 years of 24/7/365 use out of any reasonably complex integrated circuit...
Isn't 10 years close enough to forever for you? :p
Americans, especially, are totally obnoxious when it comes to languages: many of them refuse to speak any other language unless they REALLY have to
That is a completely FALSE statement. I am an American, and I absolutely REFUSE to put myself in a situation where I would be required to speak any other language.
:)
Well, it's not so doom if you are carrying your own software around, as well - to compromise, they'd need to actively scan a program that wasn't on the machine before for a decrypted private key.
I suppose it depends on your level of paranoia and how important the machine is to you. The terminal you use doesn't need anything nearly that complex. It just needs to have a keylogger to catch your passphrase, and it needs to save any files you access on your flash drive.
Also, it strikes me that it wouldn't be horrendous to construct yourself a script that would create a new keypair for you, encrypt it with a password that you only use for those (because the computer would need to have it stored), and put the private on the thumb drive and the public into authorized_keys, replacing the old one. Have it run every time that hotplug detects that particular thumb drive, and you're good. A new keypair to use from untrusted computers within by just having your thumb drive plugged in for a bit. If I didn't have a laptop to keep my keys, programs, etc. on, I'd probably go ahead and implement that now...
You'd possibly use OPIE instead, like I suggested to Basje. I am not using it myself, so there may be security implications I am currently unaware of with it. There is an OPIE password generator for Palm OS, Windows, and Unix, so you can store those on your flash drive. It is safe to use them since the password is only good once.
Or you could try the low-tech route and write a bunch of passwords down and keep changing your password in your bash_logout :). I'd rather give Opie a shot first though, since it seemed to be stable and secure already.
Is there a way to implement one time passwords with ssh?
There is an OTP package called OPIE. I toyed with it a year or so ago, it seemed to work OK. I have not used it in production, so be sure to research it first.
There is a password generator for Palm OS called pilOTP. I played around with both OPIE and pilOTP and they seemed to work quite well.
I wouldn't mind carrying around a slip of paper with 50 or so generated passwords, to login to my server. I'd cross off one at a time, each time only allowing the next one to be able to login some maintenance account.
Neat idea, but I don't think it will work with OPIE. OPIE is a challenge/response system I believe. If you do find a system that will let you do it with paper, be sure to obfuscate it somehow. If you lose the paper you want to have time to revoke those passwords.
I do not know how secure it would be but you could probably rig up your idea with a list of passwords and some trickery in your bash_logout. I do not know if I would trust it, but it might be a fun project :)