The power supply will burn itself out before it catches anything on fire. "Dust bunnies" are highly combustible but they typically don't burn long enough, only a couple of seconds, to catch anything else on fire. The real danger with the computer is probably the power strip. Make sure you spring for a good quality, UL listed power strip. The cheap ones make poor connection with the plug and use smaller-gauge wire internally and for the cord, thus increasing the risk of fire under heavy load.
A bug always seems small and stupid until someone finds a way to easily exploit it. The recent Apache bug is a great example. At first announced as unexploitable on non-Windows 32 bit systems, some freak had a ready-to-run root exploit for it in less than 3 days. The process of locating a security issue and the subsequent process of developing an exploit for it are two different processes and two different disciplines.
Unbelievable... some coworkers and I were just yesterday laughing at a resume we came across because it had the person's slashdot karma on it. Now I wonder if I should add that to my resume...
The time travel feature is odd, because the origins of postgres are from a grad student's work where the feature you describe was the entire point. He was writing an RDBMS that could service queries relative to what the database looked like at a particular time. The article I read about this didn't mention what had become of this, I figured the guy had finished it (this was like 15 years ago) and the feature had been long since removed. Maybe in reality the feature was never finished.
I think the issue with WAL piping might be in transactions. There is a huge thread somewhere on google where the primary postgres developers were starting to discuss replication and I remember WAL duplication as being one of the methods someone came up with holes in. They essentially tore holes in every method that everyone was coming up with. Then, of course, the idea they came up with was absurdly advanced. The idea of using unicast for rdbms replication is, I would say, groundbreaking.
The information in that post is flat-out wrong, there is no master-master replication for postgresql. There is barely master-slave, as in master-slave can be done via a really kludgey set of patches but there are serious caveats with regards to stored procedures.
The postgres developers (who are not responsible for any of the replication patches and have really just recently started to discuss the issue) do have some seriously good ideas with regards to replication. They plan on writing it around an existing open-source IP multicast framework. This innovation could propel postgres to the head of the RDBMS pack, however, it's "years away". Until that point, if you need more than 200-400 connections to the backend, you are going to have to either kludge or look elsewhere.
Replication is certainly a feature of the majority of the commercial RDBMSs and essentially none of the open source RDBMSs. Replication is essential in applications where you need to make more connections to the database backend than can be accomplished with one machine.
As for learning SQL, since the book specifically deals with the Sybase variants and doesn't deal with any postgresql variants it makes more sense for the book to include Sybase than Postgres.
What do you do with your postgresql database when you need to accept more than 300-400 concurrent connections to the backend? I've used postgresql for four years, extremely heavily and without issue, and don't have an answer to that question.
But yes, you are right. Postgresql is one of the most standards-compliant databases.
Postgres 7.2 implements all of the features a modern relational database is supposed to support? How about master-master replication? Even master-slave replication is only supported through third-party patches that do not scale well. The open source databases (postgres, mysql) are poorly scalable and this lack of scalability makes it impossible to even start to compare them to their commercial counterparts.
We had plenty of hardware, our software team was just expecting the end product to be able to deliver 300 concurrent connections within the 200 gb limitation. We expected to operate 7-10 machines at this load level. The CPU utilization was fine on twin 750 MHz boxes but with 2 gigs it was up in the swap after about 220 connections, at which point performance understandably dies. I could get 250 connections if I set the processes to die after 3 requests but then the fork overhead starts to kill the box.
Requests were processed at 10 per second. This is where we start getting into the postgres backend's lack of scalability.
I really feel that modperl is just a reasonably nasty hack.
I made the mistake of selecting perl as the language in which to develop a rather sizeable web based content delivery and streaming media system for use by major latin american media companies. By the time we were entrenched in the modperl-based web frontend, we discovered that 2 gigs of ram was not enough to service just 200 concurrent web connections. It's not memory leaks, per se, the processes "just grow" each request, regardless of whether or not you preload the modules into the parent. We ended up having to kill each process off after just 3 requests which really impacts performance under high request rates. I chose perl because Template Toolkit is probably the absolute best way in the universe to separate application logic from presentation markup but if I did it again I would probably use jsp.
memory utilization under modperl leaves a lot to be desired and really makes it almost unuseable for heavy duty web sites unless they are as simplistic as slashdot.
I think you are confusing light rail (trolley, streetcars) with regular rail. GM purportedly was responsible for the demise of light rail, ie replacing streetcars with buses, but this really doesn't have anything to do with regular passenger trains.
The decision by PBIA to not use the equipment had nothing to do with the accuracy of the equipment. Here is a less sensationalized story from the local newspaper which states "PBIA's decision to remove the equipment and not buy it reflects the federal government's takeover of airport security". The article mentions that the tests of the equipment were solicited right after the 9/11 incident, prior to the federal government announcing it would be addressing airport security. So the inaccuracy is not the reason this technology didn't end up in the airport.
Yeah, excellent theory. I am surprised that the investigation has not focused more in this direction. Everything I have read about this issue seems to indicate that the investigation ignores the fact that the guy has said it only happens from hotels and the presence of the hotel PBX.
As a former adminstrator of hotel telephone switches, I suspect that either the hotel's switches are being hacked into (most hotels retain default passwords) or the hotel or someone at the hotel is being paid to change the hotel switch's call routing tables. This is a simpler explanation than the telco switches being hacked and is right in line with the plaintiff's "it only happened from hotels" statement.
The issue with postgresql is the lack of "good" replication. There are currently a couple of patches for master-slave replication, although they seem to be fairly primitive. There is nothing for multi-master. I have seen indications that stuff is in the works but it will be a year or more. This kind of limits postgresql's scalability, particularly with its one-process-per-connection backend.
If Conrail doomed themselves, most programmers would probably be happy to doom themselves in a similar fashion. Conrail merged with CSX in 1998, CSX is one of the largest transportation companies in the world.
That's not how the Cher "Believe" effect was done. There was an article about it in Sound On Sound. It was done by alternating between the actual vocals and the vocals ran through a vocoder.
What you forget to mention is that the amount of CO2 given off when photosynthesis is not occurring is a mere fraction of the amount of CO2 the plant uses when the plant is photosynthesizing.
Wouldn't it be ironic to have written a song titled "Ironic" with lyrics that describe situations seemingly exemplifying irony but in fact none of the situations you described in the song were actually ironic? Hmmm...
Slashdot Mirror
The power supply will burn itself out before it catches anything on fire. "Dust bunnies" are highly combustible but they typically don't burn long enough, only a couple of seconds, to catch anything else on fire. The real danger with the computer is probably the power strip. Make sure you spring for a good quality, UL listed power strip. The cheap ones make poor connection with the plug and use smaller-gauge wire internally and for the cord, thus increasing the risk of fire under heavy load.
maru
A bug always seems small and stupid until someone finds a way to easily exploit it. The recent Apache bug is a great example. At first announced as unexploitable on non-Windows 32 bit systems, some freak had a ready-to-run root exploit for it in less than 3 days. The process of locating a security issue and the subsequent process of developing an exploit for it are two different processes and two different disciplines.
maru
Unbelievable... some coworkers and I were just yesterday laughing at a resume we came across because it had the person's slashdot karma on it. Now I wonder if I should add that to my resume...
maru
With the ultra-rare exception, I don't think the serious businessman will be reading slashdot.
maru
The time travel feature is odd, because the origins of postgres are from a grad student's work where the feature you describe was the entire point. He was writing an RDBMS that could service queries relative to what the database looked like at a particular time. The article I read about this didn't mention what had become of this, I figured the guy had finished it (this was like 15 years ago) and the feature had been long since removed. Maybe in reality the feature was never finished.
I think the issue with WAL piping might be in transactions. There is a huge thread somewhere on google where the primary postgres developers were starting to discuss replication and I remember WAL duplication as being one of the methods someone came up with holes in. They essentially tore holes in every method that everyone was coming up with. Then, of course, the idea they came up with was absurdly advanced. The idea of using unicast for rdbms replication is, I would say, groundbreaking.
maru
The information in that post is flat-out wrong, there is no master-master replication for postgresql. There is barely master-slave, as in master-slave can be done via a really kludgey set of patches but there are serious caveats with regards to stored procedures.
The postgres developers (who are not responsible for any of the replication patches and have really just recently started to discuss the issue) do have some seriously good ideas with regards to replication. They plan on writing it around an existing open-source IP multicast framework. This innovation could propel postgres to the head of the RDBMS pack, however, it's "years away". Until that point, if you need more than 200-400 connections to the backend, you are going to have to either kludge or look elsewhere.
maru
maru
Replication is certainly a feature of the majority of the commercial RDBMSs and essentially none of the open source RDBMSs. Replication is essential in applications where you need to make more connections to the database backend than can be accomplished with one machine.
As for learning SQL, since the book specifically deals with the Sybase variants and doesn't deal with any postgresql variants it makes more sense for the book to include Sybase than Postgres.
maru
maru
What do you do with your postgresql database when you need to accept more than 300-400 concurrent connections to the backend? I've used postgresql for four years, extremely heavily and without issue, and don't have an answer to that question.
But yes, you are right. Postgresql is one of the most standards-compliant databases.
maru
Postgres 7.2 implements all of the features a modern relational database is supposed to support? How about master-master replication? Even master-slave replication is only supported through third-party patches that do not scale well. The open source databases (postgres, mysql) are poorly scalable and this lack of scalability makes it impossible to even start to compare them to their commercial counterparts.
maru
We had plenty of hardware, our software team was just expecting the end product to be able to deliver 300 concurrent connections within the 200 gb limitation. We expected to operate 7-10 machines at this load level. The CPU utilization was fine on twin 750 MHz boxes but with 2 gigs it was up in the swap after about 220 connections, at which point performance understandably dies. I could get 250 connections if I set the processes to die after 3 requests but then the fork overhead starts to kill the box.
Requests were processed at 10 per second. This is where we start getting into the postgres backend's lack of scalability.
I really feel that modperl is just a reasonably nasty hack.
maru
I made the mistake of selecting perl as the language in which to develop a rather sizeable web based content delivery and streaming media system for use by major latin american media companies. By the time we were entrenched in the modperl-based web frontend, we discovered that 2 gigs of ram was not enough to service just 200 concurrent web connections. It's not memory leaks, per se, the processes "just grow" each request, regardless of whether or not you preload the modules into the parent. We ended up having to kill each process off after just 3 requests which really impacts performance under high request rates. I chose perl because Template Toolkit is probably the absolute best way in the universe to separate application logic from presentation markup but if I did it again I would probably use jsp.
maru
memory utilization under modperl leaves a lot to be desired and really makes it almost unuseable for heavy duty web sites unless they are as simplistic as slashdot.
maru
I think you are confusing light rail (trolley, streetcars) with regular rail. GM purportedly was responsible for the demise of light rail, ie replacing streetcars with buses, but this really doesn't have anything to do with regular passenger trains.
maru
What are the 8 common anti debate tactics?
The decision by PBIA to not use the equipment had nothing to do with the accuracy of the equipment. Here is a less sensationalized story from the local newspaper which states "PBIA's decision to remove the equipment and not buy it reflects the federal government's takeover of airport security". The article mentions that the tests of the equipment were solicited right after the 9/11 incident, prior to the federal government announcing it would be addressing airport security. So the inaccuracy is not the reason this technology didn't end up in the airport.
maru
Yeah, excellent theory. I am surprised that the investigation has not focused more in this direction. Everything I have read about this issue seems to indicate that the investigation ignores the fact that the guy has said it only happens from hotels and the presence of the hotel PBX.
maru
As a former adminstrator of hotel telephone switches, I suspect that either the hotel's switches are being hacked into (most hotels retain default passwords) or the hotel or someone at the hotel is being paid to change the hotel switch's call routing tables. This is a simpler explanation than the telco switches being hacked and is right in line with the plaintiff's "it only happened from hotels" statement.
maru
You are assuming that "the location" refers to the location of the event and not the location required by the ISP (perhaps a colo facility).
maru
Country music is certainly not the most popular form of music, it's one of the LEAST popular forms of music.
maru
The issue with postgresql is the lack of "good" replication. There are currently a couple of patches for master-slave replication, although they seem to be fairly primitive. There is nothing for multi-master. I have seen indications that stuff is in the works but it will be a year or more. This kind of limits postgresql's scalability, particularly with its one-process-per-connection backend.
maru
If Conrail doomed themselves, most programmers would probably be happy to doom themselves in a similar fashion. Conrail merged with CSX in 1998, CSX is one of the largest transportation companies in the world.
maru
That's not how the Cher "Believe" effect was done. There was an article about it in Sound On Sound. It was done by alternating between the actual vocals and the vocals ran through a vocoder.
maru
What you forget to mention is that the amount of CO2 given off when photosynthesis is not occurring is a mere fraction of the amount of CO2 the plant uses when the plant is photosynthesizing.
maru
Wouldn't it be ironic to have written a song titled "Ironic" with lyrics that describe situations seemingly exemplifying irony but in fact none of the situations you described in the song were actually ironic? Hmmm...
maru
That's the problem, Debian doesn't backport the bugfixes as I think you are suggesting. They backport security fixes.
I concur on the redhat issue. They are addressing security issues faster but the base installation is still not as secure as Debian.
maru