I don't get your point. There is a technical reason why different databases need the same key, yet omitting that considered a feature, not a bug. Do you mean that is is better to make integration difficult by not considering the big picture? That sounds like a make-work project.
An NI number is a reasonable unqiue primary key, although it is not guarenteed to be uniqie, and some people might not have an NI number. Because it it used for work does not preclude it from being used for other purposes.
A better solution would be to use a person-number, which is assigned at birth and lasts til you die.
Read An Introduction to Database Systems (C.J.Date), chapter 5, Data Integrity, to see why mechanisms for automatically generating a "unique" key don't work in big systems, especially when they have to be integrated into even bigger ones. In short, automatically generated "unique" keys have no inherent meaning with respect to the tuple they identify, which means that they are not systematic. This means that the same object in different databases would have different primary keys, making integration far harder because identification of related data is only possible with another reference table.
A better approach is to systematically confer keys either by using a unique combination of the fields from the data itself, or by having a single resource to confer sequence numbers unique across all systems. This is no red-herring.
> You only need a unique identifier
> within the system...
This would present problems when systems are integrated. If we use different primary keys in each system, then if we want to integrate the systems, we would need reference tables (with the associated maintenance problems) to do that.
> no need exists to identify a person
> outside the system
No need may exists when the system is first specified, but later, a need might arise which can only be fulfilled by integrating the systems, and that is when a truly unique key for each person provided dividends.
> attaching my identity at the airport to my
> identity in my real life is completely
> unnecessary to the task of travel
You imply that airports are not real-life. It may be necessary to associate your identity at the airport to your real life identity if there is a crash; if an intelligence organisation needs to profile your travel habits; if you commit a crime that is discovered later; if you loose your camera; if you die while on-board or need medication; if you want to pay in a cashless society; etc. etc. etc.
If we make an assumption that we go cashless, perhaps in 5 to 10 years, I'd need a way to establish my identity to buy anything, so a primary key number for each person might be the best thing, yes.
It is ironic that many slashdotters work in information systems, yet they are anxious about identity systems. To function, databases depend on unique primary keys for each record, and from that comes the need to overtly establish a person's identity. It isn't just in airlines where this is done. You have a number on your car for a similar reason, and a NI number of some sort. If we want to systematise, this is the trade-off, isn't it?
It's only property if you can keep it. You could keep your property with a big fucking bat, but you can't keep people from sharing data with that. If you can't keep your IP, perhaps it should be regraded from being property to something else. You can't own space or the air, or the sea, so we don't class that as property.
The yanks changed GOD to GOSH, JESUS to GEEWIZZ, TITBIT to TIDBIT , FUCKING to FREAKING , SHIT to SUGAR and all sorts of other nonsense I expect.
I am not America bashing. Everybody already knows that Americans are influenced by Christian fundamentalism, and that American society is overly prudish as a result. This should not be news to anybody, so I can't understand why you are complaining.
We have a local 'museum' for this kind of stuff. Sorting it is easy - they just mix it together with all kinds of other useless stuff, and keep it outside in a large heap.
I find it amusing that the British are now 90 % metric, but colonies like America stick with thier old system of inches and feet.
The metric system is French, and many Canadians are French. Metrication was an attempt to throw a bone to Quebec; a concession to stop them breaking up the country. So if it wasn't for Quebec, Canada would be stuck on the old Britsh colonial system in the same way that the US is.
The page has a comments section - I've posted the following. Does anybody else have views on this aspect?
Given that global warming is a bigger threat to mankind than terrorism, shouldn't we be reducing the amount of stuff we carry around in cars? Do you care at all about the air pollution you cause by carting all that iron around, not to mention the power it uses up?
MS project is the standard tool people use for making Gantt charts, levelling resources and sending them around on emails. It's the done thing in Project Management circles, and everybody expects MS project to be there. Other stuff won't do - it has to be MS Project because they have loaded it with little gizmos that hook people in. And it works OK'ish.
That type of programming is not as big a part of the sector as it once was. I assume it's some kind of engineering number crunching. I'd expect C(++) to do very well if you lay out the rows, columns and loops properly and optimise it to prefetch the data into cache memory and parallelise the instructions across the chip pipelines. Fortran might do better even.
Java does not strike me as the best kind of language for that type of work. If you use objects, the memory access could be jumbled and prefetch/pipelining might not be effective. If, however, you use arrays of primitive data types in Java, maybe the prefetch performance could be OK, but I wonder whether Java could arrange instruction pipelining?
I knew Oracle ran on a lot of hardware types, and I have wondered how they arranged this. It occurred to me that if they only develop on Linux, how can they get it to work on other things? Presumably this porting engineering group have all manner of hardware with which to roll out the different binaries. Is there a full list of all the platform support? Do they offer everything on all supported platforms?
It gives you a common installation procedure - that has some benefits. First, Oracle only has to maintain one version of the installation software, so they save time/costs and risks. Second, DBAs have a machine transparent installation procedure, so (in theory) they don't need separate training on installing on Unix, NT, Linux etc. In actual fact, there are some system level tweaks that are different.
I what way doesn't it work? There is usually a need to set up DBA and Oracle users and groups and set up some Kernel parameters before the install, but what is wrong with that - any other things? Unless you mean the first Linux release (8.1.7/8i?), which I admit was a pain to install.
This is the natural thing to do. Oracle started out on VMS and Unix type systems, and departed later into Windows. Since they ported their install process to Java between 8.1.6 and 8.1.7, and with their moves into the Application Server arena, it is clear that they have platform transparency in mind. Coupled with the fact that Unix is the dominant server platform, and Linux is a decent form of free Unix, this is a good move.
Slashdot Mirror
I don't get your point. There is a technical reason why different databases need the same key, yet omitting that considered a feature, not a bug. Do you mean that is is better to make integration difficult by not considering the big picture? That sounds like a make-work project.
An NI number is a reasonable unqiue primary key, although it is not guarenteed to be uniqie, and some people might not have an NI number. Because it it used for work does not preclude it from being used for other purposes. A better solution would be to use a person-number, which is assigned at birth and lasts til you die.
Read An Introduction to Database Systems (C.J.Date), chapter 5, Data Integrity, to see why mechanisms for automatically generating a "unique" key don't work in big systems, especially when they have to be integrated into even bigger ones. In short, automatically generated "unique" keys have no inherent meaning with respect to the tuple they identify, which means that they are not systematic. This means that the same object in different databases would have different primary keys, making integration far harder because identification of related data is only possible with another reference table. A better approach is to systematically confer keys either by using a unique combination of the fields from the data itself, or by having a single resource to confer sequence numbers unique across all systems. This is no red-herring.
> You only need a unique identifier > within the system ...
This would present problems when systems are integrated. If we use different primary keys in each system, then if we want to integrate the systems, we would need reference tables (with the associated maintenance problems) to do that.
> no need exists to identify a person
> outside the system
No need may exists when the system is first specified, but later, a need might arise which can only be fulfilled by integrating the systems, and that is when a truly unique key for each person provided dividends.
> attaching my identity at the airport to my
> identity in my real life is completely
> unnecessary to the task of travel
You imply that airports are not real-life. It may be necessary to associate your identity at the airport to your real life identity if there is a crash; if an intelligence organisation needs to profile your travel habits; if you commit a crime that is discovered later; if you loose your camera; if you die while on-board or need medication; if you want to pay in a cashless society; etc. etc. etc.
If we make an assumption that we go cashless, perhaps in 5 to 10 years, I'd need a way to establish my identity to buy anything, so a primary key number for each person might be the best thing, yes.
It is ironic that many slashdotters work in information systems, yet they are anxious about identity systems. To function, databases depend on unique primary keys for each record, and from that comes the need to overtly establish a person's identity. It isn't just in airlines where this is done. You have a number on your car for a similar reason, and a NI number of some sort. If we want to systematise, this is the trade-off, isn't it?
Actually, the Scottish got to the moon first. See Armstrong Clan. They Scots ride bikes prettry good, too.
It's only property if you can keep it. You could keep your property with a big fucking bat, but you can't keep people from sharing data with that. If you can't keep your IP, perhaps it should be regraded from being property to something else. You can't own space or the air, or the sea, so we don't class that as property.
It would be odd if America put up an iron curtain so soon after the one in Europe fell.
Fair do's, but only Colonel Blimps have ever used that sort of language!
I'm surprised you couldn't work out that this was a typo.
I am not America bashing. Everybody already knows that Americans are influenced by Christian fundamentalism, and that American society is overly prudish as a result. This should not be news to anybody, so I can't understand why you are complaining.
Get this - Americans changed the English word TITBIT to TIDBID because the original was too obscene!!!
We have a local 'museum' for this kind of stuff. Sorting it is easy - they just mix it together with all kinds of other useless stuff, and keep it outside in a large heap.
The metric system is French, and many Canadians are French. Metrication was an attempt to throw a bone to Quebec; a concession to stop them breaking up the country. So if it wasn't for Quebec, Canada would be stuck on the old Britsh colonial system in the same way that the US is.
"Usually by the uneducated" is not even a sentence, so I don't know how you have the nerve to chip in!
You seem pretty sure of yourself, so how about this for a simple answer - obey the law or risk getting sued - it's up to you.
Given that global warming is a bigger threat to mankind than terrorism, shouldn't we be reducing the amount of stuff we carry around in cars? Do you care at all about the air pollution you cause by carting all that iron around, not to mention the power it uses up?
The simple answer is that you shouldn't copy material - it's against the law. Manufactures like me have to make a living, for heaven's sake.
Why should anybody refrain from voicing thier personal political opinion? I don't get it - what do you suggest?
MS project is the standard tool people use for making Gantt charts, levelling resources and sending them around on emails. It's the done thing in Project Management circles, and everybody expects MS project to be there. Other stuff won't do - it has to be MS Project because they have loaded it with little gizmos that hook people in. And it works OK'ish.
That type of programming is not as big a part of the sector as it once was. I assume it's some kind of engineering number crunching. I'd expect C(++) to do very well if you lay out the rows, columns and loops properly and optimise it to prefetch the data into cache memory and parallelise the instructions across the chip pipelines. Fortran might do better even. Java does not strike me as the best kind of language for that type of work. If you use objects, the memory access could be jumbled and prefetch/pipelining might not be effective. If, however, you use arrays of primitive data types in Java, maybe the prefetch performance could be OK, but I wonder whether Java could arrange instruction pipelining?
I knew Oracle ran on a lot of hardware types, and I have wondered how they arranged this. It occurred to me that if they only develop on Linux, how can they get it to work on other things? Presumably this porting engineering group have all manner of hardware with which to roll out the different binaries. Is there a full list of all the platform support? Do they offer everything on all supported platforms?
It gives you a common installation procedure - that has some benefits. First, Oracle only has to maintain one version of the installation software, so they save time/costs and risks. Second, DBAs have a machine transparent installation procedure, so (in theory) they don't need separate training on installing on Unix, NT, Linux etc. In actual fact, there are some system level tweaks that are different. I what way doesn't it work? There is usually a need to set up DBA and Oracle users and groups and set up some Kernel parameters before the install, but what is wrong with that - any other things? Unless you mean the first Linux release (8.1.7/8i?), which I admit was a pain to install.
This is the natural thing to do. Oracle started out on VMS and Unix type systems, and departed later into Windows. Since they ported their install process to Java between 8.1.6 and 8.1.7, and with their moves into the Application Server arena, it is clear that they have platform transparency in mind. Coupled with the fact that Unix is the dominant server platform, and Linux is a decent form of free Unix, this is a good move.