Poor wording by grandparent. Apple always kept to the letter of the LGPL -- they dumped source when they released binaries. One big chunk of undercommented source.
Nine months ago, that changed when Apple started exposing the VCS for WebKit and actively helping kHTML developers to integrate the WebKit improvements to KDE, and integrate the newer kHTML code to WebKit. In the initial situation, Apple benefited. After opening the repository, everybody benefited, and now Apple is saying "Thank you," in a very tangible way.
Odds are that one web developer who was assigned to make the change had to document that it was fully tested in compliance with (insert management buzzword here). In fact, because of Sarbanes-Oxley, some companies lawyers have gone a little overboard with documentation requirements, and it might take a month or two just to get the appropriate signatures to allow that change. It's absolutely ridiculous, but if he doesn't have that documentation, he risks getting himself fired. Mockery will not change that simple fact, just make suicide more likely.
As far as I understand, that "rule" went out with the easy availability of proportional type. Real typesetting has never used that rule. I freely admit that I've been fully indoctrinated and can't end a bloody sentence (or line of code) without a double space.
They're planning something like that, I think, but done at the hardware level, not architecture-independent. The Intel chips that are coming out all support VT (Virtualization Technology, IIRC). Basically, think hardware-assisted VMware. Excellent. It actually has a chance of doing what IBM and Apple were touting back in the Old Days(tm), with the Pink microkernel-based OS that would have AIX and MacOS running on the same kernel, simultaneously on the same box, like IBM's VM on the big iron.
I don't think you understand anything about processors, chipsets and motherboards. Switching to AMD would require a completely different motherboard design.
I don't think you get the point. It's not hardware. It's software and compatibility. If Apple were to ship a Xserve with a pair of Opteron 280 chips (drool), it would work with the same software as the Intel-based boxes. Apple would need to write a few drivers for it. (Oh, wait. Darwin already runs on AMD hardware. Never mind.)
From the hardware design perspective, it's big. It's hard work. But from the customer's perspective, it will work the same with the same software and without another transition and another architecture compiled in the fat^H^H^universal binaries.
'Course, they probably sold their collective soul to Intel, at least for the next few years.
Despite the belief of many on/., you probably don't want a cell. It doesn't do out-of-order execution. Unless code is seriously optimized for the exact micro-architecture of the chip, probably hand-optimized for critical portions, you will get horrible performance.
In that respect, it's quite similar to the Itanium (no hardware branch prediction, all in compiler) -- screaming fast for something that's very well optimized, but change the processor (Itanic II), and you get bad performance on code compiled for the first rev of the chip.
Now, for the specific case of Photoshop, the cell might work quite well -- as a coprocessor, for filters and other ops in Photoshop, but not to run the main UI. Same thing for something like Mathematica or Matlab. It's not a very good general-purpose core (in terms of being easy to program or easy to get good performance).
It will show in the PS3 -- the first games will have horrible performance compared to games that come out six months later, as the developers understand the oddities of the Cell better and the tools get better. Yes, that difference shows up in every console that comes out, but I think it's going to be especially pronounced on the PS3 because of the cell.
No, shared code is stored in segments "owned" by each process, with copy-on-write semantics. An OS can handle this with the appropriate hardware support. The parent's ideas aren't quite right, but they don't preclude use of shared libraries at all.
It's damned near free to build wifi once you actually rebuild the infrastructure you're talking about. More to the point, the city needs to get a tax base working again. Something like this will get people and businesses to move back to the city while having a very small marginal cost (when you lay new glass for phone service, and fix the electrical grid, the cost of adding a couple of strands of glass and a power drop to a new box on the light pole is extremely small, and it's all the infrastructure cost that's needed for wifi). IIRC, the networking hardware (routers, APs, etc.) are being donated, so the overall marginal cost of adding wifi to the city now, while rebuilding is very small. Adding it after rebuilding would be much more expensive.
Most of the cost of deploying something like city-wide wifi is infrastructure -- you need to lay physical wires (glass, usually) to a large number of locations and build a box to put the hardware in, along with supplying power to that box and making it weatherproof (and tamperproof, to a reasonable degree). A Cisco AP is pocket change by comparison to those costs.
The thing about New Orleans is that they're basically starting from scratch in large parts of the city. They have to lay out new power and communications lines through large areas, and the incremental cost of an additional few strands of glass is nothing. They have to rebuild all of the traffic lights, street lights, etc. The real incremental cost of adding the infrastructure for the city-wide wifi is insignificant, and the other work needs to be done.
It has the benefit of getting people (and businesses) to come back. People that live there pay taxes. People that don't live there don't, at least not to the city. The city needs the tax base. I'm betting that someone pulled some numbers out of their arse, threw it in a spreadsheet, and showed a net fiscal gain for the city to install free wireless. Hell, they might even be right.
The key here is that it's nowhere near as expensive to install something like this for New Orleans as it would be for an undamaged city, perversely enough... just because of how much rebuilding will need to be done anyway. Best to rebuild it right.
OK. Think large university with about 40000 computers on campus on any given day. The network has gigabit pipes to the desktop, and a 20G backbone between each of about 25 core nodes. Your outbound connections to the Internet and to Internet2, and possibly other regional internets (note the lowercase 'i') have a total bandwidth over 30Gbit/s.
Some of those computers are laptops that roam from wired connection to AP 1 to another AP in a different city, but still on the institution's network. Our example roaming gnome with laptop also has access to 47 different UNIX systems and a couple of Windows terminal servers, where his communications could originate (and there are 5000 other people who have access to the same systems.
Now, you are a central net admin. You now receive a subpoena requiring all traffic generated by user A on the network. Anywhere on the network.
Complying with that order sure as hell isn't as simple as "a linux box with two NICs could do this transparently." It requires a huge amount of infrastructure, especially since CALEA requires them to do this without notice to the user, so there's no running to his office and dropping another box in front of his, not to mention that you don't want to give the FBI all the traffic from those multiple-user UNIX systems and Windows terminal servers -- only the traffic this "person of interest" is generating.
All of these numbers are reasonably close to actual for my employer, the University of Minnesota, who I sure as hell do *not* speak for in this or any other post to/. It's not a joke, and it's not an exaggeration. The problem is that big, and that expensive.
That's pretty f*cked up right there. The specs for the box clearly state that it will work with four drives. From the Sun store page listing entry-level servers:
Sun Fire X4100 Server
Fast, reliable, and energy-efficient x64 server that runs Solaris OS, Linux, and Windows.
Up to 2 dual- or single-core AMD Opteron 200 Series processors
Up to 16 GB of memory
Up to two disk drives with DVD-ROM or up to four disk drives without DVD-ROM
1U high form factor
I don't know why it won't let us actually configure it that way. And yes, only four 73G drives is pretty weak.
Quick explanation -- think of it as a chroot on steroids, or a slight enhancement over freebsd jails. It's a virtual userland, with a separate init that shares the same kernel and devices. Since the init for a zone is just another process on the system, it can have its memory, CPU, and bandwidth usage managed by the Solaris SRM tools.
You can set it up so that zone A gets a guarantee of 80% of the physical memory, and 90% of the CPU -- run your production site there. The root zone (where you manage other zones) gets a guarantee of 10% of physical memory and 5% of the CPU available. At this point, you can run zone B, with a testing site or development environment that has a guarantee of 10% of the memory and 5% of the CPU. Where it gets cool is when you realize that if zone A only needs 60% of the memory and 40% of the CPU at any given time, zone B now has 40% of the memory and probably about 60% of the CPU available to it.
It's like an extremely dynamic version of VMware that only runs Solaris. The closest LInux analog is Xen, but Linux doesn't have the resource management tools that make this really work.
There are a few other reasons as well - there are commercial applications and environments that don't support Linux (though not anywhere near as many as there used to be). More importantly, the Java VM on Solaris is better than the Linux port. If you're running a J2EE system ($DEITY help you) or a simple servlet engine, I've had better luck on Solaris than with Linux. Of course, I could probably give you just as many reasons to go the other way...
No, both the 4200 and 4100 can handle 4 SAS disks. The difference is that the 1U box, the 4100, can only handle 4 disks if you *don't* install an optical drive. Generally, in the Sun world, there's not much use for an optical anyway, and in the rare case that it's "needed," you can boot from an external USB.
My favorite bad ad of all time was from Subway. Now, Subway has had some bad ads around, but this one was just wonderful.
The ad was on the radio, and it was for some chipotle-chicken sub or something, and it had this little jingle that just went, "Chipotle! Chipotle! da da da, Chipotle!"
The problem came in when I didn't realize that it was an ad for Subway. I thought it was an ad for Chipotle, not Subway.
Ummm.... the Sun 4100 is a 1U, 4-core box -- available now. It can be stacked with 16G of memory and up to four disks (2.5" SAS [Serial Attached Storage]), or two disks and an optical drive.
The 2100 is basically a reference design -- the 4100 and 4200 boxes are designed entirely by Sun and are most excellent.
How come the gov hasnt set up a database driven site to enter names of those rescued, nationally available, so that folks can find each other?
It's simple. Government (and gov't employees) don't move that fast.
I can only assume that you're talking about the Feds, because the state governments involved don't have the infrastructure to do much anymore. Here's the secret -- the Feds don't employ many programmers. It's all outside contractors. It would take weeks just to get the bidding process for such a project done.
Also, having the rescuers enter the names in this type of database takes time away from the more important work -- rescuing the large numbers of people that are still fooked.
I'm as cynical as the next guy about large companies, probably more, but $3 or more per gallon is entirely reasonable in a normal supply and demand situation. The refineries in this country have been running at almost 100% to meet demand as it was, and 20% of the refining capacity for the nation is down. That's close to 100% for the affected region. That's going to increase prices, probably without increasing profits.
I expect to see shortages and rationing within two weeks if the refineries stay offline (due to significant damage). Yeah, it's great that W decided to tap the strategic reserves -- but those reserves, if I remember correctly, are heavy crude, not refined petrol. It's an excellent move politically, and it does help with a major problem -- panic -- but it's not going to help with refined petroleum prices or availability.
Cell phones can work on different carriers networks, they have to for 911 calls.
This is true -- in an extremely limited sense. In the US, every major carrier traditionally used different technology. AT&T/Cingular Orange uses TDMA. AT&T/Cingular Blue (their newer network), and T-Mobile use GSM (the same as the rest of the world). Verizon uses CDMA, Sprint uses PCS (a CDMA variant running on a different frequency). Most Verizon, Sprint and Cingular Orange phones can fall back to AMPS -- the old analog phone network.
Within a given technology, the phones can go between carriers. Smaller carriers like Midwest Wireless use the same technologies as the big boys, and you can roam between networks, but there's no possible way for a CDMA phone to work on a GSM network.
Usually there's only one provider for a given technology in a geographic area -- the rest will have roaming agreements to use that network. Even for a phone with the same technology, there are impediments. Most phones sold in the US are provider-locked in some way. Even if I have a phone from T-Mobile, I can't use a Cingular Blue phone number (SIM card) with it.
I'm not disputing your general point -- that wireless service or free cell calls are not the most useful donation right now. However, in the context of the article, it's what T-Mobile has to give.
Slashdot Mirror
Nine months ago, that changed when Apple started exposing the VCS for WebKit and actively helping kHTML developers to integrate the WebKit improvements to KDE, and integrate the newer kHTML code to WebKit. In the initial situation, Apple benefited. After opening the repository, everybody benefited, and now Apple is saying "Thank you," in a very tangible way.
Odds are that one web developer who was assigned to make the change had to document that it was fully tested in compliance with (insert management buzzword here). In fact, because of Sarbanes-Oxley, some companies lawyers have gone a little overboard with documentation requirements, and it might take a month or two just to get the appropriate signatures to allow that change. It's absolutely ridiculous, but if he doesn't have that documentation, he risks getting himself fired. Mockery will not change that simple fact, just make suicide more likely.
As far as I understand, that "rule" went out with the easy availability of proportional type. Real typesetting has never used that rule. I freely admit that I've been fully indoctrinated and can't end a bloody sentence (or line of code) without a double space.
They're planning something like that, I think, but done at the hardware level, not architecture-independent. The Intel chips that are coming out all support VT (Virtualization Technology, IIRC). Basically, think hardware-assisted VMware. Excellent. It actually has a chance of doing what IBM and Apple were touting back in the Old Days(tm), with the Pink microkernel-based OS that would have AIX and MacOS running on the same kernel, simultaneously on the same box, like IBM's VM on the big iron.
From the hardware design perspective, it's big. It's hard work. But from the customer's perspective, it will work the same with the same software and without another transition and another architecture compiled in the fat^H^H^universal binaries.
'Course, they probably sold their collective soul to Intel, at least for the next few years.
In that respect, it's quite similar to the Itanium (no hardware branch prediction, all in compiler) -- screaming fast for something that's very well optimized, but change the processor (Itanic II), and you get bad performance on code compiled for the first rev of the chip.
Now, for the specific case of Photoshop, the cell might work quite well -- as a coprocessor, for filters and other ops in Photoshop, but not to run the main UI. Same thing for something like Mathematica or Matlab. It's not a very good general-purpose core (in terms of being easy to program or easy to get good performance).
It will show in the PS3 -- the first games will have horrible performance compared to games that come out six months later, as the developers understand the oddities of the Cell better and the tools get better. Yes, that difference shows up in every console that comes out, but I think it's going to be especially pronounced on the PS3 because of the cell.
No, shared code is stored in segments "owned" by each process, with copy-on-write semantics. An OS can handle this with the appropriate hardware support. The parent's ideas aren't quite right, but they don't preclude use of shared libraries at all.
That's a USB hub, a mouse, and a memory stick to Mr. Computer.
I've seen worse marketese, but not much.
It's damned near free to build wifi once you actually rebuild the infrastructure you're talking about. More to the point, the city needs to get a tax base working again. Something like this will get people and businesses to move back to the city while having a very small marginal cost (when you lay new glass for phone service, and fix the electrical grid, the cost of adding a couple of strands of glass and a power drop to a new box on the light pole is extremely small, and it's all the infrastructure cost that's needed for wifi). IIRC, the networking hardware (routers, APs, etc.) are being donated, so the overall marginal cost of adding wifi to the city now, while rebuilding is very small. Adding it after rebuilding would be much more expensive.
The thing about New Orleans is that they're basically starting from scratch in large parts of the city. They have to lay out new power and communications lines through large areas, and the incremental cost of an additional few strands of glass is nothing. They have to rebuild all of the traffic lights, street lights, etc. The real incremental cost of adding the infrastructure for the city-wide wifi is insignificant, and the other work needs to be done.
It has the benefit of getting people (and businesses) to come back. People that live there pay taxes. People that don't live there don't, at least not to the city. The city needs the tax base. I'm betting that someone pulled some numbers out of their arse, threw it in a spreadsheet, and showed a net fiscal gain for the city to install free wireless. Hell, they might even be right.
The key here is that it's nowhere near as expensive to install something like this for New Orleans as it would be for an undamaged city, perversely enough... just because of how much rebuilding will need to be done anyway. Best to rebuild it right.
Some of those computers are laptops that roam from wired connection to AP 1 to another AP in a different city, but still on the institution's network. Our example roaming gnome with laptop also has access to 47 different UNIX systems and a couple of Windows terminal servers, where his communications could originate (and there are 5000 other people who have access to the same systems.
Now, you are a central net admin. You now receive a subpoena requiring all traffic generated by user A on the network. Anywhere on the network.
Complying with that order sure as hell isn't as simple as "a linux box with two NICs could do this transparently." It requires a huge amount of infrastructure, especially since CALEA requires them to do this without notice to the user, so there's no running to his office and dropping another box in front of his, not to mention that you don't want to give the FBI all the traffic from those multiple-user UNIX systems and Windows terminal servers -- only the traffic this "person of interest" is generating.
All of these numbers are reasonably close to actual for my employer, the University of Minnesota, who I sure as hell do *not* speak for in this or any other post to /. It's not a joke, and it's not an exaggeration. The problem is that big, and that expensive.
I guess we're all going back to the BBS... Dammit.
In that case, the "damage" the Internet will route around is the United States. Simple.
Quick explanation -- think of it as a chroot on steroids, or a slight enhancement over freebsd jails. It's a virtual userland, with a separate init that shares the same kernel and devices. Since the init for a zone is just another process on the system, it can have its memory, CPU, and bandwidth usage managed by the Solaris SRM tools.
You can set it up so that zone A gets a guarantee of 80% of the physical memory, and 90% of the CPU -- run your production site there. The root zone (where you manage other zones) gets a guarantee of 10% of physical memory and 5% of the CPU available. At this point, you can run zone B, with a testing site or development environment that has a guarantee of 10% of the memory and 5% of the CPU. Where it gets cool is when you realize that if zone A only needs 60% of the memory and 40% of the CPU at any given time, zone B now has 40% of the memory and probably about 60% of the CPU available to it.
It's like an extremely dynamic version of VMware that only runs Solaris. The closest LInux analog is Xen, but Linux doesn't have the resource management tools that make this really work.
There are a few other reasons as well - there are commercial applications and environments that don't support Linux (though not anywhere near as many as there used to be). More importantly, the Java VM on Solaris is better than the Linux port. If you're running a J2EE system ($DEITY help you) or a simple servlet engine, I've had better luck on Solaris than with Linux. Of course, I could probably give you just as many reasons to go the other way...
Network installs are the way of the future. ;)
The ad was on the radio, and it was for some chipotle-chicken sub or something, and it had this little jingle that just went, "Chipotle! Chipotle! da da da, Chipotle!"
The problem came in when I didn't realize that it was an ad for Subway. I thought it was an ad for Chipotle, not Subway.
These aren't that bad by comparison.
The 2100 is basically a reference design -- the 4100 and 4200 boxes are designed entirely by Sun and are most excellent.
The Sun Opteron boxes are also Microsoft certified for Windows 2003. That had to make McNealy chafe.
C. Whatever direction got me away fastest.
I can only assume that you're talking about the Feds, because the state governments involved don't have the infrastructure to do much anymore. Here's the secret -- the Feds don't employ many programmers. It's all outside contractors. It would take weeks just to get the bidding process for such a project done.
Also, having the rescuers enter the names in this type of database takes time away from the more important work -- rescuing the large numbers of people that are still fooked.
I'm as cynical as the next guy about large companies, probably more, but $3 or more per gallon is entirely reasonable in a normal supply and demand situation. The refineries in this country have been running at almost 100% to meet demand as it was, and 20% of the refining capacity for the nation is down. That's close to 100% for the affected region. That's going to increase prices, probably without increasing profits.
I expect to see shortages and rationing within two weeks if the refineries stay offline (due to significant damage). Yeah, it's great that W decided to tap the strategic reserves -- but those reserves, if I remember correctly, are heavy crude, not refined petrol. It's an excellent move politically, and it does help with a major problem -- panic -- but it's not going to help with refined petroleum prices or availability.
Within a given technology, the phones can go between carriers. Smaller carriers like Midwest Wireless use the same technologies as the big boys, and you can roam between networks, but there's no possible way for a CDMA phone to work on a GSM network.
Usually there's only one provider for a given technology in a geographic area -- the rest will have roaming agreements to use that network. Even for a phone with the same technology, there are impediments. Most phones sold in the US are provider-locked in some way. Even if I have a phone from T-Mobile, I can't use a Cingular Blue phone number (SIM card) with it.
I'm not disputing your general point -- that wireless service or free cell calls are not the most useful donation right now. However, in the context of the article, it's what T-Mobile has to give.
</pedant>
Actually, only twenty. And that's with a pair of optional SCSI backplanes (Sun 530-2744) that are a pain-in-the-arse to install.
;)
Still, quite a bit more disk than a stock PC.