There is one problem with your argument. Most, if not all, new car dealers are franchises. A number of automakers (such as Saturn AFAIK) do require that their dealers sell no other brands of new car. Some others, such as Subaru, don't appear to care. However, for many, it is part of the franchising agreement.
As franchises, they are required to follow the rules that the franchiser sets. However, PC manufacturers aren't franchisers. If the "Intel Inside" stuff is voluntary, then I don't see a problem wtih it. Manufacturers don't have to pay the fee in exchange for advertising. While there are a number of brand loyalists, a lot of people, especially when shopping for fairly basic workstations (corporate) or lower end home machines, don't care.
Intel is making a lot of restrictions for participation, such as requiring that non-Intel machines be sold under a different brand or sub brand. While this may seem extreme, it makes sense. The cost of adding an additional label (i.e Compaq Presario, IBM Aptiva) is small, and lets you sub-brand while still taking advantage of the well known parent brand. Intel doesn't want to pay to advertise for non-intel machines. Note that a sub-brand is no big deal. Most major PC manufacturers have several already.
Personally, I'm quite happy with the performance of my Intel CPU. I felt it to be the best chip for the job, given that I run a lot of floating point intensive applications, but need Linux/ia32 compatability. If I didn't care about FP performance, I probably would have bought a K6.
PSI.net, which provides the White House's link to the outside world, did not immediately return phone calls late Sunday. Neither did a White House spokesman.
From everything I've seen, PSI is one of the worst backbone providers in terms of quality of service, and reliability. I wouldn't put it past PSI to have an outage that they don't manage to fix for 24hr or more, that affects the whitehouse.gov connection (or more likely, co-located machine).
Let's figure the bandwidth of a station wagon of magnetic tape. We'll use DLT, since that's the current technology in tapes. To be specific, we will use a DLT-IV tape that holds 35GB uncompressed. We will assume no compression. According to Quantum's datasheet, a DLT-IV tape is 4.16 by 4.16 by 1 inch. A small to midsize station wagon gives you 4 feet by 4 feet by 3 feet. (i.e. Volvo 240 since I know the approximate measurements). This means that the said wagon can hold 11 by 11 by 36 tapes, for a total of 4356 tapes. This is only using the trunk, and ignoring the seats. This is also ignoring the possibility of packing a few extras on the side using the 'slack' space. These 4356 tapes hold a total of 152460 gigabytes. According to mapsonus.com, going from New York to Los Angeles by car takes 56 hours 34 minutes, which is 203640 seconds. This means that a small station wagon going from New York to Los Angeles, and with only the trunk filled with DLT, has a bandwidth of approximately 6 gigabits per second. Extrapolate this to an 18-wheeler and you can see that even the fastest fiber optic data lines have a long way to go to even come close to a truck of DLT.
From what the article says, Redhat is waiting for at least a draft version of the standard before saying they will support it. They feel that standards are overhead (they are, to a degree).
There are two separate issues here: the value of standards, and when to adopt a standard.
Standards are what the internet (as we know it) is built on. The Web would not have happened without a standardized http, nor compatability between so many e-mail systems without SMTP. Standards like HTTP, FTP, Telnet, TCP/IP, and DNS are what have made the internet possible. What happens if an application decides to handle DNS differently than everybody else? It won't work. What about a web browser that sends nonstandard HTTP queris? It won't work, at least not consistently. Having said that, standards can be overhead. I'm sure there are flaws in all of the accepted internet standards, but they work, and they work well.
On the other hand, there are bad standards. It seems that most of these are produced by groups that don't regularly handle standardization (you don't see too many bad ones from IEEE, IETF, etc). The LSB is a committee, and might or might not do a good job. Redhat's wait-and-see attitude towards the standard is quite reasonable.
Actually, according to the Jargon file, FUBAR was one of the best cases of getting jargon past suits. On a VAX, FUBAR stood for Failed UniBus Address Register. In fact, if your VAX was suffering from a FUBAR, then it was probably FUBAR.
While this is from experience with the local telco to Redhat (GTE), a lot of this is conjecture. It'd be interesting to hear from somebody there what really happened once they're back online. When doing a server move like this, you're also moving your circuit. RedHat has a DS-3, which is a lot more work for the telco to move than a DS-1. What they would have done would be to coordinate the physical move and having GTE move the DS-3 (GTE is the only DS-3 loop provider in their area AFAIK) at the same time. Due to IP issues it can't be active at both ends at once. What probably happened was that Redhat went down, moved their machines, and found out that GTE screwed up with the circuit. AFAIK they have a single DS-3 and no redundancy, which is pretty common, especially since they're on the GTE RTP SONET ring, so additional circuits would all come off of the same SONET, so they wouldn't get much of any real redundancy anyway. GTE is very well known for being late in installing things. (I'm from that area. I worked for an ISP a few minutes from Redhat last summer.)
Modern commercial tape technology, specifically DLT, has gotten very fast and reliable. While I realize most of you haven't dealt with anything larger than a peecee and therefore find real technology hard to deal with, it is out there. A Quantum DLT7000 drive, for under $5k, can write 35GB native (70GB compressed; onboard hardware compression) at 5MB/sec media speed. Also, according to Quantum's specs, a DLT cart has a storage life of "More than 30 year with less than 5% loss in demagnetization (at 20C and 40% non-condensing humidity)". DLT is very fast, reliable, reasonably priced (given what it does), and has been around for a while. If they're using DAT (or other helical scan technology) for all this data, they need to get their head(s) checked.
Slashdot Mirror
There is one problem with your argument. Most, if not all, new car dealers are franchises. A number of automakers (such as Saturn AFAIK) do require that their dealers sell no other brands of new car. Some others, such as Subaru, don't appear to care. However, for many, it is part of the franchising agreement.
As franchises, they are required to follow the rules that the franchiser sets. However, PC manufacturers aren't franchisers. If the "Intel Inside" stuff is voluntary, then I don't see a problem wtih it. Manufacturers don't have to pay the fee in exchange for advertising. While there are a number of brand loyalists, a lot of people, especially when shopping for fairly basic workstations (corporate) or lower end home machines, don't care.
Intel is making a lot of restrictions for participation, such as requiring that non-Intel machines be sold under a different brand or sub brand. While this may seem extreme, it makes sense. The cost of adding an additional label (i.e Compaq Presario, IBM Aptiva) is small, and lets you sub-brand while still taking advantage of the well known parent brand. Intel doesn't want to pay to advertise for non-intel machines. Note that a sub-brand is no big deal. Most major PC manufacturers have several already.
Personally, I'm quite happy with the performance of my Intel CPU. I felt it to be the best chip for the job, given that I run a lot of floating point intensive applications, but need Linux/ia32 compatability. If I didn't care about FP performance, I probably would have bought a K6.
From the wired article:
PSI.net, which provides the White House's link to the outside world, did not immediately return phone calls late Sunday. Neither did a White House spokesman.
From everything I've seen, PSI is one of the worst backbone providers in terms of quality of service, and reliability. I wouldn't put it past PSI to have an outage that they don't manage to fix for 24hr or more, that affects the whitehouse.gov connection (or more likely, co-located machine).
Let's figure the bandwidth of a station wagon of magnetic tape. We'll use DLT, since that's the current technology in tapes. To be specific, we will use a DLT-IV tape that holds 35GB uncompressed. We will assume no compression. According to Quantum's datasheet, a DLT-IV tape is 4.16 by 4.16 by 1 inch. A small to midsize station wagon gives you 4 feet by 4 feet by 3 feet. (i.e. Volvo 240 since I know the approximate measurements). This means that the said wagon can hold 11 by 11 by 36 tapes, for a total of 4356 tapes. This is only using the trunk, and ignoring the seats. This is also ignoring the possibility of packing a few extras on the side using the 'slack' space. These 4356 tapes hold a total of 152460 gigabytes. According to mapsonus.com, going from New York to Los Angeles by car takes 56 hours 34 minutes, which is 203640 seconds. This means that a small station wagon going from New York to Los Angeles, and with only the trunk filled with DLT, has a bandwidth of approximately 6 gigabits per second. Extrapolate this to an 18-wheeler and you can see that even the fastest fiber optic data lines have a long way to go to even come close to a truck of DLT.
There are two separate issues here: the value of standards, and when to adopt a standard.
Standards are what the internet (as we know it) is built on. The Web would not have happened without a standardized http, nor compatability between so many e-mail systems without SMTP. Standards like HTTP, FTP, Telnet, TCP/IP, and DNS are what have made the internet possible. What happens if an application decides to handle DNS differently than everybody else? It won't work. What about a web browser that sends nonstandard HTTP queris? It won't work, at least not consistently. Having said that, standards can be overhead. I'm sure there are flaws in all of the accepted internet standards, but they work, and they work well.
On the other hand, there are bad standards. It seems that most of these are produced by groups that don't regularly handle standardization (you don't see too many bad ones from IEEE, IETF, etc). The LSB is a committee, and might or might not do a good job. Redhat's wait-and-see attitude towards the standard is quite reasonable.
Actually, according to the Jargon file, FUBAR was one of the best cases of getting jargon past suits. On a VAX, FUBAR stood for Failed UniBus Address Register. In fact, if your VAX was suffering from a FUBAR, then it was probably FUBAR.
Before posting ignorant comments about the lack of availability of truetype fonts for X, take a look at xfstt.
I don't have the URL offhand, but it exists;
this Netscape looks far better with it than
without out.
While this is from experience with the local telco to Redhat (GTE), a lot of this is conjecture. It'd be interesting to hear from somebody there what really happened once they're back online. When doing a server move like this, you're also moving your circuit. RedHat has a DS-3, which is a lot more work for the telco to move than a DS-1. What they would have done would be to coordinate the physical move and having GTE move the DS-3 (GTE is the only DS-3 loop provider in their area AFAIK) at the same time. Due to IP issues it can't be active at both ends at once. What probably happened was that Redhat went down, moved their machines, and found out that GTE screwed up with the circuit. AFAIK they have a single DS-3 and no redundancy, which is pretty common, especially since they're on the GTE RTP SONET ring, so additional circuits would all come off of the same SONET, so they wouldn't get much of any real redundancy anyway. GTE is very well known for being late in installing things. (I'm from that area. I worked for an ISP a few minutes from Redhat last summer.)
Modern commercial tape technology, specifically DLT, has gotten very fast and reliable. While I realize most of you haven't dealt with anything larger than a peecee and therefore find real technology hard to deal with, it is out there. A Quantum DLT7000 drive, for under $5k, can write 35GB native (70GB compressed; onboard hardware compression) at 5MB/sec media speed. Also, according to Quantum's specs, a DLT cart has a storage life of "More than 30 year with less than 5% loss in demagnetization (at 20C and 40% non-condensing humidity)". DLT is very fast, reliable, reasonably priced (given what it does), and has been around for a while. If they're using DAT (or other helical scan technology) for all this data, they need to get their head(s) checked.