Seriously, I'd like someone to explain what precisely about Oracle could ever be considered absolutely necessary that cannot be found anywhere else aside from organizational bias and insipid politics.
Here are a few answers:
1) Oracle allows you to tune types of transactions very heavily. For example some tables or transaction types can log while other do not.
2) Log miner combined with archiver allow you to generate activity reports offline
3) Complex partition tables with partition indexes
4) Grid databases so you can spread a consistent database over hundreds of servers
5) Really good ISAM compatibility on their mainframe versions so that legacy cobol code will work against semi-relational structures
Bit by bit how the data is laid out is covered in Oracle literature. Database concepts chapter 2 covers blacks and 5 covers the rows (the data). There really isn't anything stopping you from writing a perl script to pull data out. Of course Oracle has all sorts of easy export routines so no one bothers.
I think you should read the context. The debate was whether the article could be considered "anti-Stallman". I.E. it was in the context of why it did or didn't matter if he choose thee GPL given his previous stances. My point was the attacks were generalized.
In the mid 1980s (really until about 1993) floating point operations were very expensive. This choice seems stupid now that floats are first class objects. But most languages today have vector operations in libraries even though most modern processors can handle them natively.
n this situation I prefer to throw an Exception. I've written the method to return a float but I've been asked to evaluate a negative number. This does not evaluate to a float.
My example is a little simple to make it clear why this isn't such a good idea. Basically using your method (called function checks) you are often testing the same criteria N times. Lets say you have a function like:
A which will end up calling B which end up calling C which will end up calling D. D has criteria (x) shared by C,B and A. Then: calling function checks (x), A checks (x), B checks (x), C checks (x), D checks (x). Now if its something like
x = (y >=0) then that's pretty cheap. But lets say it was much more complicated. Lets say for example a Neural Net with a million elements where you need to make sure the points are convex in 25,000 dimensional space.
That is if you have to test this as a function without global knowledge you have to: 1) convert each point into a 25k element vector 2) take all ordered triples of points (10^18) of them 3) do a simple computation on each triple of vectors. Something like 400k simple math operations. Total time on a modern pentium IV should work out a over a millisecond per triple which works out to a little over 30,000 years. Maybe you can improve and cut it to 1000 years if you are very good at assembly.
conversely if you check the points as they come in and can assure A,B,C and D of that you skip it. Otherwise you have to do this test over and over and over every-time you call function A. The situation doesn't have to be nearly as bas as the neural net for it not to be a good idea to be checking this over and over and over.
My 'contract' is the mathematical definition of SQRT,
Which mathematical definition? For example is (-3) a valid sqrt for 9? If the calling function only allows for the positive sqrt and the sqrt function returns either then you'll get flagged and know to call absolute value?
These were used extensively where DTO's (Data Transport Object) provided the data on which the method acted. Yes, there was quite a bit of verification code in the methods to assure the DTO was complete (especially with Weblogic XMLObjects), and that might be simpler to do given a contractual-based language.So my question is: isn't these the same thing?
No in the contract based language your weblogic routine would specify what a complete DTO meant and the caller would be responsible for making sure it met your criteria. You would never check. Eiffel would assume the lower level code between client and server checks to make sure the transport package is complete. That is you wouldn't have code dealing with network (did the database I/O complete) and code dealing with business logic at the same level. The database code would assure a complete transmission and the form code would know the original DTO was perfect so....
Well Java is from a different generation, its harder to compare. But C++ compiles native. Most early Eiffels compiled to C, and Meyers explicitly consider C to be the "p-code" for Eiffel (15.1.2). In 15.4.2 he outlines the higher level calling mechanism. I don't generally respond to ACs but you need to defend this point a little better.
He doesn't talk about the NRA issue in section 11. He address that in "5. A SKEWED MORAL PERSPECTIVE". He mentions this again in item 6 from your list. Anyway I don't see what assertion specifically you are refuting.
You aren't remembering correctly, he went after all open source and free software. His biggest criticism of Stallman and Linus was for not distancing the free software movement from the NRA when Raymond tried to link the two (he held Stallman more responsible since Linus being a Finn might not have known what the NRA was...). He attacked FreeBSD explicitly for poor quality.
What is funny though he lists a whole bunch of reasons vendors turn something open source that pretty much amount to "it isn't worth keeping as commercial product for reason X" where he lists the possible X's. Anyway here is a reminder:
The big thing is design by contract. Calling procedures specify what they will guarantee functions and what they expect, functions guarantee what they won't change and list what they expect. This makes code reuse actually practical, since global effects are spelled out.
For example a sqrt function will require x >= 0, that is the calling function must have checked. Since this is specifically listed in the contract your calling function is expected to be able to indicated how it knows that x >= 0 before computing sqrt(x).
The second big thing is that the language is pretty high level and you just give hints to the compiler about how to implement things (sort of like Oracle SQL) unlike Java or C++ where implementation details are definitely part of the program.
OK and the problem with that is what? $1500 per programmer per year is not a large expense. It would likely go much lower if TrollTech started getting huge volumes.
RedHat -- pretty clearly was against KDE for licensing reasons. Has since the last 1990 been trying to maintain good relations with the KDE people. Continues to offer a version of KDE as a standard that is integrated with RedHat.
Ubuntu - - has a Kubuntu project on relatively equal footing
Suse -- You are just dead wrong on this one. Here is policy
There has been a lot of debate over the last few days in the media and on message boards about what is happening regarding GNOME vs. KDE at Novell. Here's the situation. All future enterprise-class Linux product releases, including Novell Linux Desktop, SUSE Linux Enterprise Server and Novell Open Enterprise Server, will continue to ship with both the GNOME and KDE desktop environments. In upcoming versions of Novell enterprise applications, the default desktop environment will be GNOME. When customers install Novell Linux products, they will be given the option to choose either the GNOME or KDE environment during the installation process. If the user makes no explicit choice, GNOME will be installed.
This change has no implication for current Novell customers. Novell will continue to invest in both GNOME and KDE and we will continue to offer maintenance and support for these products and their desktop environments throughout their planned product lifetimes.
This decision has very little impact on either the openSUSE project or future versions of SUSE Linux. SUSE Linux will continue to showcase the newest open source desktop, development, and server technologies in a complete, stabilized operating system. SUSE Linux will continue to deliver both the GNOME and the KDE desktop environments along with the latest open source packages for those environments.
The only distribution I know of that is openly hostile to KDE (i.e. will not support it) is UserLinux and as a result it was rejected almost instantly as a standard by the Linux community.
Seems his post had content. Doesn't like the hidden options, doesn't like the color schemes (or lack thereof), doesn't like lack of configurability.... Yours had none.
Most people don't really object to bloat in user space. Lots of people like the Apple model where most libraries are static and bundled with the apps and thus on the hard-drive many times.
But even if there were a desire to get rid of bloat I don't think your solution would work. Embedded programmers don't tend to have one very limited function and not think too much about expansion into related tasks. So you would end up with multiple commands which do similar kinds of things with only slight differences. Further they don't share libraries since they need slightly different libraries. So at every level you'll end up with code/apps which make use of callouts to very inflexible systems and then glue them together. Sort of like the COBOL apps of the 1960s-70s.
And why should MS continue to develop Mac Office? For $125 (student/teacher) you can buy the Windows OS and then run the PC version of Office. The same is true for all the other "fringe" software; just add $200 to cost and bundle it with the Windows OS, then every software vendor can claim Mac compatibility.
1) They can claim it all they want. Mac users are very picky about design and don't like the design of mac apps. 2) This is a dual boot not VM solution. Having to reboot to another OS is a large burdon 3) They have been claiming this for years via. virtual PC. The mac crowd doesn't buy it.
Actually you are still doing a bit more. Yeah that would work fine (though still a lot more CPU intensive but no extra disk). From the description it didn't sound like there was a "thumbs down" button if there is then I agree I'm picking up a guy who hates the things you like. But then again why does he keep going to the same sites?
That's too expensive. You are doing the dot product against all 10m users. Using my system you are only hitting a small subset of the users (those that are in the same blocks as you). Also the computation of a sum is much much cheaper than a full dot product.
Obviously a weighted comparison works best but even if you compute that monthly you are talking 3 trillion dot products per day.
Slashdot Mirror
Then the VAR is abusing the embedded license. If the customer knows they are running Oracle they can't offer that level of support.
Seriously, I'd like someone to explain what precisely about Oracle could ever be considered absolutely necessary that cannot be found anywhere else aside from organizational bias and insipid politics.
Here are a few answers:
1) Oracle allows you to tune types of transactions very heavily. For example some tables or transaction types can log while other do not.
2) Log miner combined with archiver allow you to generate activity reports offline
3) Complex partition tables with partition indexes
4) Grid databases so you can spread a consistent database over hundreds of servers
5) Really good ISAM compatibility on their mainframe versions so that legacy cobol code will work against semi-relational structures
Bit by bit how the data is laid out is covered in Oracle literature. Database concepts chapter 2 covers blacks and 5 covers the rows (the data). There really isn't anything stopping you from writing a perl script to pull data out. Of course Oracle has all sorts of easy export routines so no one bothers.
Good stocks rarely make good companies and vice versa. Apple is selling for 40x earnings, 4.5x sales and 7x book. Its a little on the high side.
I think you should read the context. The debate was whether the article could be considered "anti-Stallman". I.E. it was in the context of why it did or didn't matter if he choose thee GPL given his previous stances. My point was the attacks were generalized.
In the mid 1980s (really until about 1993) floating point operations were very expensive. This choice seems stupid now that floats are first class objects. But most languages today have vector operations in libraries even though most modern processors can handle them natively.
Its the Republicans that allowed the return of bribery to American politics.
n this situation I prefer to throw an Exception. I've written the method to return a float but I've been asked to evaluate a negative number. This does not evaluate to a float.
My example is a little simple to make it clear why this isn't such a good idea. Basically using your method (called function checks) you are often testing the same criteria N times. Lets say you have a function like:
A which will end up calling B which end up calling C which will end up calling D. D has criteria (x) shared by C,B and A. Then:
calling function checks (x), A checks (x), B checks (x), C checks (x), D checks (x). Now if its something like
x = (y >=0) then that's pretty cheap. But lets say it was much more complicated. Lets say for example a Neural Net with a million elements where you need to make sure the points are convex in 25,000 dimensional space.
That is if you have to test this as a function without global knowledge you have to:
1) convert each point into a 25k element vector
2) take all ordered triples of points (10^18) of them
3) do a simple computation on each triple of vectors. Something like 400k simple math operations. Total time on a modern pentium IV should work out a over a millisecond per triple which works out to a little over 30,000 years. Maybe you can improve and cut it to 1000 years if you are very good at assembly.
conversely if you check the points as they come in and can assure A,B,C and D of that you skip it. Otherwise you have to do this test over and over and over every-time you call function A. The situation doesn't have to be nearly as bas as the neural net for it not to be a good idea to be checking this over and over and over.
My 'contract' is the mathematical definition of SQRT,
Which mathematical definition? For example is (-3) a valid sqrt for 9? If the calling function only allows for the positive sqrt and the sqrt function returns either then you'll get flagged and know to call absolute value?
These were used extensively where DTO's (Data Transport Object) provided the data on which the method acted. Yes, there was quite a bit of verification code in the methods to assure the DTO was complete (especially with Weblogic XMLObjects), and that might be simpler to do given a contractual-based language.So my question is: isn't these the same thing?
No in the contract based language your weblogic routine would specify what a complete DTO meant and the caller would be responsible for making sure it met your criteria. You would never check. Eiffel would assume the lower level code between client and server checks to make sure the transport package is complete. That is you wouldn't have code dealing with network (did the database I/O complete) and code dealing with business logic at the same level. The database code would assure a complete transmission and the form code would know the original DTO was perfect so....
I said biggest criticism of Stallman, not biggest assertion. The post was one paragraph.
Well Java is from a different generation, its harder to compare. But C++ compiles native. Most early Eiffels compiled to C, and Meyers explicitly consider C to be the "p-code" for Eiffel (15.1.2). In 15.4.2 he outlines the higher level calling mechanism. I don't generally respond to ACs but you need to defend this point a little better.
He doesn't talk about the NRA issue in section 11. He address that in "5. A SKEWED MORAL PERSPECTIVE". He mentions this again in item 6 from your list. Anyway I don't see what assertion specifically you are refuting.
You aren't remembering correctly, he went after all open source and free software. His biggest criticism of Stallman and Linus was for not distancing the free software movement from the NRA when Raymond tried to link the two (he held Stallman more responsible since Linus being a Finn might not have known what the NRA was...). He attacked FreeBSD explicitly for poor quality.
r e.html
What is funny though he lists a whole bunch of reasons vendors turn something open source that pretty much amount to "it isn't worth keeping as commercial product for reason X" where he lists the possible X's. Anyway here is a reminder:
http://www.apostate.com/programming/bm-freesoftwa
The big thing is design by contract. Calling procedures specify what they will guarantee functions and what they expect, functions guarantee what they won't change and list what they expect. This makes code reuse actually practical, since global effects are spelled out.
For example a sqrt function will require x >= 0, that is the calling function must have checked. Since this is specifically listed in the contract your calling function is expected to be able to indicated how it knows that x >= 0 before computing sqrt(x).
The second big thing is that the language is pretty high level and you just give hints to the compiler about how to implement things (sort of like Oracle SQL) unlike Java or C++ where implementation details are definitely part of the program.
That's pretty much it.
But then the IT can shift burden of proof. User would need a source that shows that they can't do function Y in format Z. This would only be one step.
State governments have fairly large IT and IS departments. Its not like they can't handle minor problems as they arise.
LSB has been around for 8 years or so. Freedesktop.org has always been loaded with standards people. This isn't new.
OK and the problem with that is what? $1500 per programmer per year is not a large expense. It would likely go much lower if TrollTech started getting huge volumes.
Ubuntu - - has a Kubuntu project on relatively equal footing
Suse -- You are just dead wrong on this one. Here is policyThe only distribution I know of that is openly hostile to KDE (i.e. will not support it) is UserLinux and as a result it was rejected almost instantly as a standard by the Linux community.
Seems his post had content. Doesn't like the hidden options, doesn't like the color schemes (or lack thereof), doesn't like lack of configurability.... Yours had none.
My favorite was vigor, "all the features twice the bugs" which took VI and added the vigor assistant
"You have attempted to move left please press confirm or cancel" http://vigor.sourceforge.net/screenshots/
Most people don't really object to bloat in user space. Lots of people like the Apple model where most libraries are static and bundled with the apps and thus on the hard-drive many times.
But even if there were a desire to get rid of bloat I don't think your solution would work. Embedded programmers don't tend to have one very limited function and not think too much about expansion into related tasks. So you would end up with multiple commands which do similar kinds of things with only slight differences. Further they don't share libraries since they need slightly different libraries. So at every level you'll end up with code/apps which make use of callouts to very inflexible systems and then glue them together. Sort of like the COBOL apps of the 1960s-70s.
Total code size would probably be larger.
I think you are forgetting the context here. The "library" being referred to by the GP is code from the Linux kernel which is pure GPL.
And why should MS continue to develop Mac Office? For $125 (student/teacher) you can buy the Windows OS and then run the PC version of Office. The same is true for all the other "fringe" software; just add $200 to cost and bundle it with the Windows OS, then every software vendor can claim Mac compatibility.
1) They can claim it all they want. Mac users are very picky about design and don't like the design of mac apps.
2) This is a dual boot not VM solution. Having to reboot to another OS is a large burdon
3) They have been claiming this for years via. virtual PC. The mac crowd doesn't buy it.
Actually you are still doing a bit more. Yeah that would work fine (though still a lot more CPU intensive but no extra disk). From the description it didn't sound like there was a "thumbs down" button if there is then I agree I'm picking up a guy who hates the things you like. But then again why does he keep going to the same sites?
That's too expensive. You are doing the dot product against all 10m users. Using my system you are only hitting a small subset of the users (those that are in the same blocks as you). Also the computation of a sum is much much cheaper than a full dot product.
Obviously a weighted comparison works best but even if you compute that monthly you are talking 3 trillion dot products per day.