I work for a company that OEMs a product with a published API. They make quarterly updates, but here's the rub...
The updates continually update their back-end in non-backwards compatible ways. We end up running multi-cpu days of regression tests to find what's broke and then spend oodles of man-days tracking down what happened and figuring out workarounds each time we try to update. We're still using the API libraries that are many versions before the latest because of this.
At one point I couldn't figure out how to do something with their API,so I requested example code. They sent part of the source a real product that they market that does what I needed. I soon discovered that they don't use their own published interface, completely bypassing the API classes entierly to get to functionality I can't.
[sarcasm]This kind of posturing is nothing new and it's wonderful to see how people can still post responses so rationally.[/sarcasm]
The open source plan put in place by HP is quite refreshing. They have a reasonable time frame to replace all the proprietary pieces with open source ones to get it all out there. They have embraced the homebrew community and made them part of the open-source direction and I wish them well.
Personally, I'm excited about the proposition. WebOS still comes in top in customer satisfaction polls, imagine that.
I have had similar "invention" agreements from all my employers. Their language seems to infer that working for them is the incubation that will bring on new ideas, so even if you're off-the-clock, it is because you are working for them that you came up with the idea at all. However, in the agreements is a request for things that you have/are working on so they will be exempted. I usually include a several page list of things that I've thought about, generic enough to cover almost any field outside of my day to day work.
That said, if you come up with an idea not related to your tasks, they would be very hard pressed to make a case against you. If you come up with a better widget than the one you're doing their, they have a good case.
See my comments above on how to spoof a supported hp printer for your non-supported one. More information on the "print to any printer" thread found on http://forums.precentral.net/
I got around this by using my server and putting three option statements in net-snmp to spoof a "supported" hp printer and then redirecting printer communication to my non-hp printer instead. Took all of 5 minutes. I've got 5 WebOS devices in my house that are heavily used. Curse you HP!
Shades of the 80s! Canon Research created a great little c-like interpreted language called ici. It had all sorts of nifty lisp like features and had a nice API for native extensions. They expected to put it in all of their products (including printers) and even open-sourced it. Outside of a few external projects that I and others had, I don't think it went anywhere.
Set your mail server to relay through your ISP. Most dynamic addresses are blocked via RBLs anyway. Unless your ISP provides reverse DNS of your address a good percentage of your mail will not be delivered. I've been doing this way before they started blocking the ports.
I would think that you would find a better discusion on the mobiread forum, since this is a forum for serious ebook readers. Personally, I would never call a backlit screen useful for serious reading. It doesn't work well in brightly lighted areas (such as outdoors), and the flicker causes eyestrain and fatigue. Reflective technologies such as eInk can be read for hours (like a book). Touch screens do reduce the contrast ratio for these type of screens.
There were a lot of studies done a long time ago, and there are some very accurate psycho-visual computer models of the human visual system. I had the pleasure of working with the Jeff Lubin model when I worked at Sarnoff Corp, which won an Emmy Award back in 2000.
The 30 fps requirement is not a fixed point, but depends upon a lot of other factors, including viewing distance, field of view, and lighting conditions. The reason that film operates at 24 fps is because it is expected to be viewed in a darkened room. When film is trans-coded for TVs, they have to modify the gamma for a normally lighted viewing area or it will look bad. NTSC TVs are interlaced, displaying 60 fields per second, even though the frame rate is 30 frames per second.
Bottom line is that this article should include the environmental factors under which this point was made.
The RCA 1800 family had an interesting feature... Every instruction took 8 clock cycles (with a few extended ones that took 12). This means that it was easy to, not only figure out how long a routine would take to run, but write routines to take up a specified time.
In the RCA VIP (hobby) computer, the tape output was a frequency-shift keyed output that was driven by a software UART created by this technique. You could also program it to play music producing the appropriate frequency using a software timing loop. I wrote a few programs that did just the opposite... did frequency decomposition using the built-in zero-crossing detector for the tape interface (a poor man's spectrum analyzer).
It was pretty amazing what could be done it 8k bytes of memory.:)
I developed a 300 baud modem for the (then) brand new RCA data terminal back around 1980. The trickiest part of this thing was that the filters had to be designed and made from discrete components.
The plant manager saw my design which had precision components and he had a fit. His instructions were, "Use the components we have in the stock room". He also saw the line transformer and he said, "This is the age of semiconductors, get rid of that transformer".
I spent weeks researching how to replace that transformer with a semiconductor device that wouldn't distort the signal, operate bilaterally (in and out), and withstand the the 10,000 volt pulse that the FCC required for testing. I finally found such a beast, so I gave the manager a choice... Use the 6 cent transformer, or replace it with a $40 semiconductor.
My complaints about using non-precision components fell on deaf ears, so I figured the only way to prove that this wouldn't work would be to build a prototype and show that it would fail. To my dismay the first prototype worked flawlessly. I was told to build 10 more for FCC testing but not a single one of these worked even though the frequency response curves were all on the mark. I finally discovered that the phase response was a mess. I finally convinced them to use precision components.
I don't work for Sun either, but I agree with many of your points. My homebrew NAS server has been running OpenSolaris since build 49 and ZFS has uncovered issues like a flaky sata cable, and an unstable power supply without losing a single bit of data. In some other systems with similar problems, I found silently corrupted data on drives with ntfs, xfs, and hfs+. I would not go to a non-zfs flle system without some kicking and screaming.
I don't know if OpenSolaris, EON, or NextentaOS is missing any of the features of FreeNAS, but I would think about these as a viable replacement. One advantage is that they will always have the bleeding edge zfs enhancements first, like the new deduplication feature.
Zfs has saved my data enough times to warrant it as an essential part of any NAS system I use and value it's contents. I'm seriously looking at the OpenSolaris based NextentaOS and EON as the only remaining choice for an open source NAS solution. I'm also using Zfs on my Mac OS/X machines after I silently lost data on both NTFS, and HFS+ drives and await the community effort to bring zfs on mac up to the latest codebase.
I've been an audiophile/engineer for a long time. In my world, there are a few "golden" ears I've met that hit things absolutely on the nose every time. They can tell the level of dynamic compression, frequency shaping, etc. One particular one is a sought-after final mix-down recording engineer. Calling audiophiles names serves no purpose other than showing you're out of the scene.
CDs are notoriously flawed. Remember that it was trying to fit a particular classical piece on a CD that caused the trade-off that came up with an 8-bit sample and a 44.1k/sec sample rate. Both of these noticeably degrade the sound. In vinyl there was a problem with SNR and tracking that caused the RIAA phono equalization. Bottom line is that both vinyl and CDs have their own colorization to deal with. Personally, I find the vinyl sound more pleasing (note I didn't say better) which has a longer serious listening period before I get fatigued (tired of listening).
You can always tell the true audiophiles because they are skeptical about the hype that comes down the pipe (remember "Holographic sound processors?). They do true engineering/subjective testing in a controlled environment to see what is real or not. The high-end audio cable is such a point. As an engineer, it was obvious that the only difference in wires are their impedance, resistance, and capacitance values. Since they are combined to produce any frequency coloration, dropping the resistance low enough will preclude reasonable values of the others from affecting the sound (within hearing range). A blind A-B test of various wires showed that any wire 14-gauge or higher was indistinguishable in my setup. Guess what I use for speaker wire... The fat zip-cord that is used for outdoor low-voltage lighting. I got it at 8 cents/foot.
The same argument goes for transistor vs. tube amplifiers. You can make a transistor amp sound exactly like a tube amplifier with a single-ended instead of a totem-pole design, and they stay in the optimum operation region a lot longer (75 years) than it's tube counterpart (6 months). Check out single-ended DIY amplifier designs at the PassLabs web site.
Bottom line, Anyone can tell the difference between lossy and non-lossy compressed audio once the loss gets big enough. What is big enough depends upon a lot of factors. In this case, I have a real problem with the equipment and choice of subjects. We like what we're familiar with, not necessarily what is better. After driving the same car for years and then moving to a different model, it usually takes some time to get adjusted to the sound system until it feels comfortable just because it's different.
The main places where coloration occurs is in the conversion of sound . That's why I built my own hybrid transmission-line/electrostatic loudspeakers. No one that has heard them has not made a comment that they heard things in familiar pieces that they didn't even know existed (a flute in the background, etc.). This is true with vinyl or CD as a source. http://www.genashor.com/gallery3/index.php/ElectroStatic-LoudSpeakers
On most systems that support ZFS, there is already the capability to to encryption via the loop back interface. They are working on zfs native encryption and it has been under test internal in Sun for awhile. De-dup capabilities will be available soon. The development of zfs has been incredibly fast for a file system. Considering the Mac implementation is at rev 8 and I've got 14 on my OpenSolaris system should say something... Apple hasn't really put much effort into keeping up with the state-of-the-art of ZFS.
I've never had corruption issues on ZFS over the several years of heavy use, ups crashes, lightning strikes, etc. During the same time, I've seen several XFS, NTFS, and HFS+ filesystems in adjacent servers get twisted beyond repair.
I've known a "mentalist" that part of his act was to reveal a person's SSN, which he did with amazing accuracy. He's been doing this trick for over 25 years.
He uses a formula from a person's birthplace and age to get most of the digits. He used other cues for the remaining digits, but I'm fuzzy about this process.
No particular reason, other than I use the same directory to back up the windows machines as well. I'm sure that the OpenSolaris NFS implementation would be fast or faster than CIFS. If you're using Linux or BSD I've heard that NFS has some occasional hiccups (stale lock issues) and performance problems. Take the previous sentence as pure gossip as I haven't seen or done any testing to prove/disprove it.
You're right on the button. I created a sparse file for each machine image using diskutil so I could fix maximum size (I'd hate it to take over my entire 2.5 TB pool). The trick is to figure out the name that each machine wants, but worse comes to worse, you cancel it quick on the first sync if it's wrong and then rename the file and start it again.
Then I used the native CIFS service that comes with OpenSolaris for the connection. I started with Samba, but the native CIFS service had 1000X better throughput.
There is an option that enables mounting "foreign" disks for time machine. This may explain it better:
Well, I've been using ZFS for several years on several different machines with mixes of mirrored and RAID-Z configurations. Since that time, I've never lost one bit of data. It has survived power-supply failures, lightning strikes that fried the motherboard, flaky I/O cards, and human error. I understand that the implementation on Mac OS/X may be buggy, but it's not inherent to ZFS.
I've several Macs doing time-machine to networked ZFS drives. It's definitely the filesystem I'd like to have everywhere.
Many high security establishments, both government and commercial, realize that they can't stop technology without serious concessions. What some do are to put a special tamper-proof tag over the camera. Then they just inspect the tag when you exit and, if tampered, confiscate the device until it can be validated.
I had to work on a bug on some critical software on a small military plane. I had to drive through a bad snowstorm with the car full of equipment for a full day to get to the plane. The plane came in late and ended up losing it's hanger space.
I was stuck on a metal chair outside the plane on the tarmac, in sub-freezing weather, using my computer for warmth. The power was being powered by a diesel generator that made thinking near impossible.
I was told that, if I didn't have it fixed by the end of the day, I would have to fix it on the way to their assignment... in Bosnia! I worked late in the evening, freezing and without breaks, but finally got things straightened out.
I hope to never go through something like that again.
The CPU should support VT extensions and have plenty of RAM. If you figure 1-2 GB of RAM for each VM running concurrently, you should have a good estimate of what you need.
A good OS platform to use is OpenSolaris because it's open source. In addition, it supports Xen, Zones, and other virtualization approaches. With OpenSolaris's Crossbow (network virtualization) you can not only have a testbed for VMs, but you can stress-test your VMs under different network characteristics (long latencies, etc.). Throw in Dtrace to track down issues and you have a real winner.
Slashdot Mirror
I work for a company that OEMs a product with a published API. They make quarterly updates, but here's the rub...
The updates continually update their back-end in non-backwards compatible ways. We end up running multi-cpu days of regression tests to find what's broke and then spend oodles of man-days tracking down what happened and figuring out workarounds each time we try to update. We're still using the API libraries that are many versions before the latest because of this.
At one point I couldn't figure out how to do something with their API,so I requested example code. They sent part of the source a real product that they market that does what I needed. I soon discovered that they don't use their own published interface, completely bypassing the API classes entierly to get to functionality I can't.
I'd take open source over this pain any day.
[sarcasm]This kind of posturing is nothing new and it's wonderful to see how people can still post responses so rationally.[/sarcasm]
The open source plan put in place by HP is quite refreshing. They have a reasonable time frame to replace all the proprietary pieces with open source ones to get it all out there. They have embraced the homebrew community and made them part of the open-source direction and I wish them well.
Personally, I'm excited about the proposition. WebOS still comes in top in customer satisfaction polls, imagine that.
I have had similar "invention" agreements from all my employers. Their language seems to infer that working for them is the incubation that will bring on new ideas, so even if you're off-the-clock, it is because you are working for them that you came up with the idea at all. However, in the agreements is a request for things that you have/are working on so they will be exempted. I usually include a several page list of things that I've thought about, generic enough to cover almost any field outside of my day to day work.
That said, if you come up with an idea not related to your tasks, they would be very hard pressed to make a case against you. If you come up with a better widget than the one you're doing their, they have a good case.
See my comments above on how to spoof a supported hp printer for your non-supported one. More information on the "print to any printer" thread found on http://forums.precentral.net/
I got around this by using my server and putting three option statements in net-snmp to spoof a "supported" hp printer and then redirecting printer communication to my non-hp printer instead. Took all of 5 minutes. I've got 5 WebOS devices in my house that are heavily used. Curse you HP!
Shades of the 80s! Canon Research created a great little c-like interpreted language called ici. It had all sorts of nifty lisp like features and had a nice API for native extensions. They expected to put it in all of their products (including printers) and even open-sourced it. Outside of a few external projects that I and others had, I don't think it went anywhere.
Set your mail server to relay through your ISP. Most dynamic addresses are blocked via RBLs anyway. Unless your ISP provides reverse DNS of your address a good percentage of your mail will not be delivered. I've been doing this way before they started blocking the ports.
And Apple would never stop you from putting your app on the App Store simply because it steps on their plans to control their platform, right?
I would think that you would find a better discusion on the mobiread forum, since this is a forum for serious ebook readers. Personally, I would never call a backlit screen useful for serious reading. It doesn't work well in brightly lighted areas (such as outdoors), and the flicker causes eyestrain and fatigue. Reflective technologies such as eInk can be read for hours (like a book). Touch screens do reduce the contrast ratio for these type of screens.
There were a lot of studies done a long time ago, and there are some very accurate psycho-visual computer models of the human visual system. I had the pleasure of working with the Jeff Lubin model when I worked at Sarnoff Corp, which won an Emmy Award back in 2000.
The 30 fps requirement is not a fixed point, but depends upon a lot of other factors, including viewing distance, field of view, and lighting conditions. The reason that film operates at 24 fps is because it is expected to be viewed in a darkened room. When film is trans-coded for TVs, they have to modify the gamma for a normally lighted viewing area or it will look bad. NTSC TVs are interlaced, displaying 60 fields per second, even though the frame rate is 30 frames per second.
Bottom line is that this article should include the environmental factors under which this point was made.
Which reminds me...
The RCA 1800 family had an interesting feature... Every instruction took 8 clock cycles (with a few extended ones that took 12). This means that it was easy to, not only figure out how long a routine would take to run, but write routines to take up a specified time.
In the RCA VIP (hobby) computer, the tape output was a frequency-shift keyed output that was driven by a software UART created by this technique. You could also program it to play music producing the appropriate frequency using a software timing loop. I wrote a few programs that did just the opposite... did frequency decomposition using the built-in zero-crossing detector for the tape interface (a poor man's spectrum analyzer).
It was pretty amazing what could be done it 8k bytes of memory. :)
Yes. The FCC required that any device connected to the POTS line could withstand a lightning strike that created a 10kV pulse down the line.
I developed a 300 baud modem for the (then) brand new RCA data terminal back around 1980. The trickiest part of this thing was that the filters had to be designed and made from discrete components.
The plant manager saw my design which had precision components and he had a fit. His instructions were, "Use the components we have in the stock room". He also saw the line transformer and he said, "This is the age of semiconductors, get rid of that transformer".
I spent weeks researching how to replace that transformer with a semiconductor device that wouldn't distort the signal, operate bilaterally (in and out), and withstand the the 10,000 volt pulse that the FCC required for testing. I finally found such a beast, so I gave the manager a choice... Use the 6 cent transformer, or replace it with a $40 semiconductor.
My complaints about using non-precision components fell on deaf ears, so I figured the only way to prove that this wouldn't work would be to build a prototype and show that it would fail. To my dismay the first prototype worked flawlessly. I was told to build 10 more for FCC testing but not a single one of these worked even though the frequency response curves were all on the mark. I finally discovered that the phase response was a mess. I finally convinced them to use precision components.
I don't work for Sun either, but I agree with many of your points. My homebrew NAS server has been running OpenSolaris since build 49 and ZFS has uncovered issues like a flaky sata cable, and an unstable power supply without losing a single bit of data. In some other systems with similar problems, I found silently corrupted data on drives with ntfs, xfs, and hfs+. I would not go to a non-zfs flle system without some kicking and screaming.
I don't know if OpenSolaris, EON, or NextentaOS is missing any of the features of FreeNAS, but I would think about these as a viable replacement. One advantage is that they will always have the bleeding edge zfs enhancements first, like the new deduplication feature.
Zfs has saved my data enough times to warrant it as an essential part of any NAS system I use and value it's contents. I'm seriously looking at the OpenSolaris based NextentaOS and EON as the only remaining choice for an open source NAS solution. I'm also using Zfs on my Mac OS/X machines after I silently lost data on both NTFS, and HFS+ drives and await the community effort to bring zfs on mac up to the latest codebase.
Thanks, I stand corrected. It's 16-bit PCM.
I've been an audiophile/engineer for a long time. In my world, there are a few "golden" ears I've met that hit things absolutely on the nose every time. They can tell the level of dynamic compression, frequency shaping, etc. One particular one is a sought-after final mix-down recording engineer. Calling audiophiles names serves no purpose other than showing you're out of the scene.
CDs are notoriously flawed. Remember that it was trying to fit a particular classical piece on a CD that caused the trade-off that came up with an 8-bit sample and a 44.1k/sec sample rate. Both of these noticeably degrade the sound. In vinyl there was a problem with SNR and tracking that caused the RIAA phono equalization. Bottom line is that both vinyl and CDs have their own colorization to deal with. Personally, I find the vinyl sound more pleasing (note I didn't say better) which has a longer serious listening period before I get fatigued (tired of listening).
You can always tell the true audiophiles because they are skeptical about the hype that comes down the pipe (remember "Holographic sound processors?). They do true engineering/subjective testing in a controlled environment to see what is real or not. The high-end audio cable is such a point. As an engineer, it was obvious that the only difference in wires are their impedance, resistance, and capacitance values. Since they are combined to produce any frequency coloration, dropping the resistance low enough will preclude reasonable values of the others from affecting the sound (within hearing range). A blind A-B test of various wires showed that any wire 14-gauge or higher was indistinguishable in my setup. Guess what I use for speaker wire... The fat zip-cord that is used for outdoor low-voltage lighting. I got it at 8 cents/foot.
The same argument goes for transistor vs. tube amplifiers. You can make a transistor amp sound exactly like a tube amplifier with a single-ended instead of a totem-pole design, and they stay in the optimum operation region a lot longer (75 years) than it's tube counterpart (6 months). Check out single-ended DIY amplifier designs at the PassLabs web site.
Bottom line, Anyone can tell the difference between lossy and non-lossy compressed audio once the loss gets big enough. What is big enough depends upon a lot of factors. In this case, I have a real problem with the equipment and choice of subjects. We like what we're familiar with, not necessarily what is better. After driving the same car for years and then moving to a different model, it usually takes some time to get adjusted to the sound system until it feels comfortable just because it's different.
The main places where coloration occurs is in the conversion of sound . That's why I built my own hybrid transmission-line/electrostatic loudspeakers. No one that has heard them has not made a comment that they heard things in familiar pieces that they didn't even know existed (a flute in the background, etc.). This is true with vinyl or CD as a source. http://www.genashor.com/gallery3/index.php/ElectroStatic-LoudSpeakers
Gary
On most systems that support ZFS, there is already the capability to to encryption via the loop back interface. They are working on zfs native encryption and it has been under test internal in Sun for awhile. De-dup capabilities will be available soon. The development of zfs has been incredibly fast for a file system. Considering the Mac implementation is at rev 8 and I've got 14 on my OpenSolaris system should say something... Apple hasn't really put much effort into keeping up with the state-of-the-art of ZFS.
I've never had corruption issues on ZFS over the several years of heavy use, ups crashes, lightning strikes, etc. During the same time, I've seen several XFS, NTFS, and HFS+ filesystems in adjacent servers get twisted beyond repair.
I've known a "mentalist" that part of his act was to reveal a person's SSN, which he did with amazing accuracy. He's been doing this trick for over 25 years.
He uses a formula from a person's birthplace and age to get most of the digits. He used other cues for the remaining digits, but I'm fuzzy about this process.
No particular reason, other than I use the same directory to back up the windows machines as well. I'm sure that the OpenSolaris NFS implementation would be fast or faster than CIFS. If you're using Linux or BSD I've heard that NFS has some occasional hiccups (stale lock issues) and performance problems. Take the previous sentence as pure gossip as I haven't seen or done any testing to prove/disprove it.
You're right on the button. I created a sparse file for each machine image using diskutil so I could fix maximum size (I'd hate it to take over my entire 2.5 TB pool). The trick is to figure out the name that each machine wants, but worse comes to worse, you cancel it quick on the first sync if it's wrong and then rename the file and start it again.
Then I used the native CIFS service that comes with OpenSolaris for the connection. I started with Samba, but the native CIFS service had 1000X better throughput.
There is an option that enables mounting "foreign" disks for time machine. This may explain it better:
http://www.macosxhints.com/article.php?story=20080420211034137
Well, I've been using ZFS for several years on several different machines with mixes of mirrored and RAID-Z configurations. Since that time, I've never lost one bit of data. It has survived power-supply failures, lightning strikes that fried the motherboard, flaky I/O cards, and human error. I understand that the implementation on Mac OS/X may be buggy, but it's not inherent to ZFS. I've several Macs doing time-machine to networked ZFS drives. It's definitely the filesystem I'd like to have everywhere.
Many high security establishments, both government and commercial, realize that they can't stop technology without serious concessions. What some do are to put a special tamper-proof tag over the camera. Then they just inspect the tag when you exit and, if tampered, confiscate the device until it can be validated.
I had to work on a bug on some critical software on a small military plane. I had to drive through a bad snowstorm with the car full of equipment for a full day to get to the plane. The plane came in late and ended up losing it's hanger space. I was stuck on a metal chair outside the plane on the tarmac, in sub-freezing weather, using my computer for warmth. The power was being powered by a diesel generator that made thinking near impossible. I was told that, if I didn't have it fixed by the end of the day, I would have to fix it on the way to their assignment... in Bosnia! I worked late in the evening, freezing and without breaks, but finally got things straightened out. I hope to never go through something like that again.
The CPU should support VT extensions and have plenty of RAM. If you figure 1-2 GB of RAM for each VM running concurrently, you should have a good estimate of what you need. A good OS platform to use is OpenSolaris because it's open source. In addition, it supports Xen, Zones, and other virtualization approaches. With OpenSolaris's Crossbow (network virtualization) you can not only have a testbed for VMs, but you can stress-test your VMs under different network characteristics (long latencies, etc.). Throw in Dtrace to track down issues and you have a real winner.