Not to mention the occasional backend crash that can be used for denial of service possibilities....
Actually, I moved from MySQL in 2000 specifically because at the time it didn't have subquery support. I still get frustrated by performance in normalized db's and other things.
The great invention of MySQL-- the ONLY database to have Write-Only-Tables.....
Don't judge a book by its cover
on
Sun Buys MySQL
·
· Score: 1
Basically, you would be close to arguing that every program written in Visual C++ is a derivative work of Windows.
My own thinking (IANAL) is that linking *never* has anything to do with derivation. Here is why:
When a program is run through a linker along with relavent libraries, one of two things happens depending on whether static or dynamic linking is requested:
1) Either the requested libraries are loaded into memory into separate contiguous address spaces along with the original application (they share the same protected memory segment, however) or 2) The requested libraries are stripped down (non-creative process, purely functional), and packaged in the same file, but separate sections.
In both cases, what you get looks a lot like a book with a single cover. Looking at the cover, you might suspect that there is only a single copyright owner. And indeed there is. However, in this case there are separate, distinct portions within it of different authorship, which suggests that this is a compiled or collected rather than derivative work.
In a similar case, suppose I write a paper on element formation in stars. I cite important papers on the subject including the Caughlin, Fowler, Zimmerman paper and include references to specific paragraphs, suggesting that if the other work is handy to the reader, they can JMP over to that other paragraph and read it for more information. Just because I say something like "See page 3, second paragraph" doesn't make my work derivative of the other. My map is not one of non-literal copyrighted elements (as might be arguably the case in a book like "The Wind Done Gone."
Ok, so suppose instead of being published in a journal, the paper is instead published in an astrophysics anthology along with several of the papers I cite. Again, one book with one cover. And infact the publisher probably owns the copyrights for the ordering and arrangement. But again, that is not a derivative work.
The real question is whether "based on" in the GPL has the same meaning as it does in the Copyright Act (where it is *clearly* defined in such a way that linking would not qualify) and whether "aggregation" means the same thing as preparing a collected or compiled work. If that is the case, then the GPL does not control linking and you can create a wonderful proprietary application that .
Note, however, that a lot of other issues could create derivative works in software. For example, reusing graphical design aspects, or allowing other non-literal creative elements of a program to creap into the new work would imply derivation.
1) Provide an option for administrators to disable any non-strict-mode settings. I.e. have a super-strict mode that clients cannot get out of.
2) Provide a syntax in table creation to cause the CREATE TABLE to error out if the requested table type is not available.
Both of the situations I described above have to do less with lack of options in MySQL, but rather the lack of central control either from an application developer perspective (no way to *force* InnoDB tables or else not create tables at all) or from a DBI perspective (no way to *force* real data bounds checking). Hence the "options" you speak of seem like they should be *requirements* rather than options for real applications.
Suppose we are collecting information. We want to collect information as to a primary place of residence for individuals on a survey but they are allowed to decline to provide that information. When they provide the information, we populate the address tables (normalizing addresses is a *real* pain) and provide a link to the address in the individual table in the primary_residence field.
When primary_residence is NULL we *know* that this means that it is unknown. Our data model expects that for *every* respondant, there is *exactly* one primary residence. Hence enforcing this functional dependency makes a great deal of sense in any normal form.
In a similar case, suppose we store expiration dates for tax rates. NULL means "no plans to change the rate yet" in other words "will probably change at some unknown time in the future." NULL is probably more semantically correct than an infinity timestamp or other fun kludges.
In short, I see nothing in BCNF, 4NF, or 5NF definitions which excludes NULLs used in this manner (i.e. information which is presumed to exist but is unknown at the moment). If the information exists, even if it is not known, the database can and should be designed to assume that it does exist.
The other issue however is that sometimes NULL is use to represent data which does not exist.
For example, consider: create table employee ( person_id int primary key references person(id), hourly_wage numeric, monthy_salary numeric CHECK hourly_wage is not null OR monthly_salary is not null, CHECK hourly_wage is null OR monthly_salary is null );
In this case, we know that either hourly_wage or monthly_salary will be null for every row but not both.
This is dangerous for a number of reasons:
1) Yes, it breaks 3NF because NULLs representing "information does not apply" suggest that there is no functional dependency on that record. In the above example, it would be better redesign the db to avoid those NULLs. 2) It introduces semantic ambiguity into the relational model because elsewhere NULL may be used to mean "UNKNOWN."
Given that normalization definitions deal with data dependencies, I dont see how this is a normalization question.
It is a valid database design question but I would be inclined to use NULL in that table unless I expected at least one person not to have a valid SSN. In which case I would break it off.
SSN is functionally dependant on the person if it exists for the person. However, if the person does *not* have a SSN, then it is not functionally dependant for every portion of the set.
However, nothing prevents you in PostgreSQL from defining a pseudo-Primary Key with NULLS. Something like: CREATE UNIQUE INDEX foo ON bar(baz) WHERE abc IS NULL;
In that case, X, NULL can only occur once in a table. That can be useful behavior too, for example when you have a table like: CREATE TABLE tax ( coa_id INT NOT NULL REFERENCES coa(id), rate NUMERIC NOT NULL, expires DATE, );
CREATE UNIQUE INDEX foo ON tax(coa_id, expires); CREATE UNIQUE INDEX bar ON tax(coa_id) WHERE expires IS NULL;
This allows you to have one rate for each account which is expected to be valid until some unknown point in the future.
non-relational information?
on
Sun Buys MySQL
·
· Score: 1
You mean like XML? Does MySQL have native xpath capabilities now?
Or do you mean something else?
In general I have generally found that relational storage of *all* information is generally preferred...
I: Economic:
A: Renew tax credits for alternative energy development.
B: Create new form of tax credit for energy conservation and distributed development (methane digestors at dairy farms and the like)
C: Work with industry on reducing scope of patents in IT-related fields.
D: Push for reform of copyright law.
II: Criminal Justice
A: Create a grading system for prisons based on recitivism rates.
B: Create a task force to study effects of sentensing lengths on recitivism rates.
C: Work with states to lower recitivism rates by adopting optimal sentensing lengths based on above studies.
III: Domestic Security and Civil Liberties
A: Create open public forums on issues relating to domestic security. Open up general discussions on abstract security issues relating to aircraft, airports, infrastructure, and federal buildings to the public.
B: Undertake a complete survey of hydroelectric dams in this country for vulnerabilities to attack. Scenarios would include small and large amounts fo conventional low and high explosives, and also thermobaric attack profiles.
C: Undertake a study of the impact on civil liberties of various post-9/11 measures.
D: Work to create a rationed guest worker program to undermine illegal immigration.
E: Work on a controlled legalization of Marijuana to undermine drug cartels which have been ammassing enough firepower to challenge the Mexican Army.
1) Import/export of Cannibis would be prohibited, except as allowed by treaty.
F: Work towards a streamlined procedure for deporting illegal immigrants with dependant American citizen children. The children would be given US passports in the process of deportation. The passports might be applied for and then distributed through local consulates, or might be distributed prior to deportation.
Foreign Policy:
I: Iraq
A: Make it *clear* that our continued involvement in Iraq requires certain rules enforced by the Iraqi government including no official or unofficial use of any militia units by any government agencies.
B: Revoke all contracts with external security firms in Iraq until further notice.
C: Make it *clear* that our continued involvement in Iraq will not exceed our welcome by the duly elected Iraqi government.
D: Seek help from The Arab League and the European Union for peacekeeping operations in Iraq under EU and AL flags.
II: Afghanistan
A: Undertake a review of Afghanistan's Constitution and its impact on governability of the country. Provide advice on this matter to that government in the form of a declassified report.
B: Request permission from Pakistan's government to pursue insurgants across the border.
III: Israel/Iran/Syria
A: Continue Bush's policy of delaying/obstructing delivery of aid when Israel is not playing well with the Palestinians.
B: Continue pushing for a Palestinian State.
C: Declare that we stand for defending the Green Line in Israel. We have no position on attacks against IDF targets outside the Green Line.
D: State to Iran and Syria that we are prepared to give a security guarantee provided that:
1) all nuclear weapons programs are verifiably halted
2) all enrichment for fissile materials is halted
3) they help a stable Iraqi government emerge
If you define a terrorist as an individual or organization which attempts to inspire terror as a way of pushing through some political agenda, that would suggest that the US Gov't is more successful in this area than Al Qaeda....
- The war on the Islamic radicals is supposed to be temporary. Why restrict what your citizens can do permanently? Who says? Has anyone articulated a victory condition in this so-called war on terrorism? BTW, if you read Hamdi v. Rumsfeld (a few years ago), the Supreme Court acknowledged that it might not be temporary.
Yeah, man, I hate when they accurately judge my risk of an accident and prevent me from leeching off of safe drivers. Ok, so frequent drinkers are more likely to get DUI's or be involved in alcohol-related accidents than teetolalers. Does that mean that you are putting yourself at risk by sometimes havind a few drinks at a bar when you have a designated driver? Since the answer is pretty obviously "no" then this means that the actuarial data is not accurate to any given case.
On the other hand, I would think that health insurance companies might like very much to be able to use racial data for determining premiums. After all, if blacks are more likely to have poverty-related diseases, then even wealthy blacks ought to have higher health insurance rates, right?
Well, the issue is that early grain uses (which involved sprouting the grain) made it susceptible to fermentation. The reason is that the sprouting process makes the seed produce amylase which in turn converts starches into sugars.
I am not sure at all that alcoholic beverages were first created by holy men. More likely they were created by accident at first and quickly deified. Mead, Wine, and Beer all have been revered as sacred in ancient cultures.
I suppose you are right-- RMS wants people to use his software. If he releases a GPL v3.1 which removes the system libraries exception, then all of GNU becomes tied to HURD's release, sorta like Duke Nukem Forever*;-)
*I have a long-running joke that Duke Nukem Forever has been repeatedly delayed due to the lack of a suitable HURD release to play it on....
Remember Thomas Bushnell? For all the ideas that free software == free speech, Thomas Bushnell was required by Stallman to *resign* from the position of being primary HURD architect (and the project has certainly not recovered from that). Why was he asked to resign? Because he (publically) argued that the GNU Free Documentation License was not free enough to be universally consistant with Debian's Free Software Guidelines.
Is there a model in place to handle the centralized liability that this would create? I mean, right now you pay for car insurance because you control the vehicle. If the contol occurs from a computer made by GE, then if a fatal crash occurs, they might be liable. The technology is mature enough we could have self-driving cars today (heck, we could have had them nearly a decade ago!) but I suspect that liability concerns are a big part of the reason we don't.
Even if self-driving cars are safer than human-driven ones, then it seems to me that it might still be prohibitiv to buy insurance.
While I won't presume to know more than the author of that book, on the face of it, it seems like a good thing to adopt new technology, even if everyone else does the same, if for no other reason than the increased efficiency it should bring. So everyone should be moving to Vista then?
It is probably better to say that it is a good thing to review new technology for potential advantages and adopt it where it makes sense. Of all the strategies I can think of nothing is more ineffective than adopting technologies because they are new.
First, I don't dispute your main points. Outsourcing generally does add to the response time and cause difficulties in business process management even beyond what you have listed. However, targetted outsourcing can off-set some of these (and only sometimes) by bringing in experts in certain types of business management. In short in the *right* places it can increase rather than decrease available IT resources.
Now, on the balance, I think that outsourcing today is *generally* done with the wrong parts of IT and generally done with the wrong goals and hence creates far more problems than it solves, but I don't think this is inherent in the idea of outsourcing some things outside areas that a business is good at.
First, IT is not a strategic investment anymore in terms of big business decisions. It is an operational investment. In short, the technology and adoption has improved to the point that one has to invest in it for operational reasons but that these don't create strategic considerations.
Secondly, the IT department isn't going to die just because Google now offers office apps. If you want to run any non-trivial computing and communications infrastructure, you need an IT department. Databases of customer data? Accounting data? etc? Sure some of this can be outsourced, but how far do you want to take it? You still want to have some control over things like authentication services, etc.
The way I see it, the shift to SAAS simply broadens the choices for career development. It doesn't eliminate corporate IT departments but simply offers them some competition.
I have been reading a fair bit of legal analysis (IANAL) relating to the GPL v2 and have been discussing various ambiguities relating to the GPL v3 with people at the SFLC. These licenses *do* have some ambiguities (though I think they are less of an issue for the GPL v2).
The major issue for the GPL v2 is that it is not 100% clear where the boundary relating to mere aggregation is. In general it is easy to read "a work based on the original work" meaning derivative work (i.e. a transformation or adaption of the original work in the same way that a movie may be based on a book, or a sequel may be based on another book), while aggregation seems to read as a collected or compiled work, but these simple interpretations are at odds with the FSF's interpretations. I.e. dynamic or even static linking would seem to create (possibly non-literal) compilations under copyright law, not derivations even if the linker strips out unused portions (this is because that process would not be creative enough to create a *new* copyrighted work in the form of the new library code). Hence the simple reading of the GPL v2 would seem to allow one to link proprietary applications to, say, GNU Readline. This question has not been resolved in court yet.
The GPL v3 has the same issue, but adds a few more. For example, does section 7, paragraph 2 govern sections of BSD code included verbatim in a GPL v3 application? I.e. must one be allowed to change the license of a file to the GPL v3 in order to call it compatible? (Eben Moglen says "Yes" while Richard Fontana says "No"-- both are members of the SFLC and both were involved in the GPL v3 development process.)
There are also a few false ambiguities-- for example the question as to whether mere use of software inside an organization might ever one to license patents out (the relevant section of the GPL v3 only applies to explicit patent licenses), though clearly one would want to stop using software before filing patent suits due to patent retaliation clauses.
It seems to me that there is a fundamental logical problem in the analysis.
IMO, the scientific method is all about information processing. All scientists do is derive experiments to gather information and process that information. If information processing is not capable of arriving at an experimental conclusion, then this generally means that one is either processing insufficient data, invalid data, or processing valid data wrong. Either way, it means that the information processing regimen is insufficient for the science.
Now, since we assume the information processing is flawed or insufficient in a mismatch, then we have to understand that holes in information processing only tell us that our existing conclusions are wrong. Hence this is *not* falsifiable and is no different than intelligent design.
There is always benefit to having multiple paths available. Separate tools generally excel at separate tasks.
Just because H.323 was available, does this mean that the IETF should have adopted it instead of RTP/SIP? Obviously when one is largely dealing with TCP/IP networks only, the latter standard is *far* better, but H.323 can be used there too (as well as many other environments-- I would expect it should be reasonably simple to run H.323 over a cell-switched or circuit-switched network for example, without the complexity that it introduces on a packet-switched network.
My first response on hearing about the deal is "no good can come of this." The fact is-- these companies directly compete in most areas and one is *far* stronger than the other.
Apple was an unusual case because Microsoft was investing in them during the hight of the DoJ investigations of their anticompetitive behavior. My own thinking is that Microsoft was terribly afraid that Apple exiting the market would mean that they would be broken up or crippled behind very onerous consent decrees. Apple thus was actually the *stronger* partner in that negotiation because they had something that Microsoft wanted even more than they needed continued support. With Novell, it is not quite the same case.
I dunno. OfficeOpenXML vs ODF is a fairly complex and technical debate which I have generally been loath to jump into. In general, there are things I like about both standards. Programatically OOXML is a better standard in terms of transformation while ODF is a better standard in terms of automatic generation. In general, I think that OOXML is a better standard for office suites' internal formats, while ODF is a more open standard for connecting to non-XML formats. Also the patent issue of OOXML is a major cause of concern.
The problem is that people generally think that one of the two must succeed at the others' expense. Personally I would like nothing better than for both sides to be willing to push for universal open source adoption of both formats without fear of patent infringement suits.
Slashdot Mirror
Not to mention the occasional backend crash that can be used for denial of service possibilities....
Actually, I moved from MySQL in 2000 specifically because at the time it didn't have subquery support. I still get frustrated by performance in normalized db's and other things.
The great invention of MySQL-- the ONLY database to have Write-Only-Tables.....
Basically, you would be close to arguing that every program written in Visual C++ is a derivative work of Windows.
My own thinking (IANAL) is that linking *never* has anything to do with derivation. Here is why:
When a program is run through a linker along with relavent libraries, one of two things happens depending on whether static or dynamic linking is requested:
1) Either the requested libraries are loaded into memory into separate contiguous address spaces along with the original application (they share the same protected memory segment, however) or
2) The requested libraries are stripped down (non-creative process, purely functional), and packaged in the same file, but separate sections.
In both cases, what you get looks a lot like a book with a single cover. Looking at the cover, you might suspect that there is only a single copyright owner. And indeed there is. However, in this case there are separate, distinct portions within it of different authorship, which suggests that this is a compiled or collected rather than derivative work.
In a similar case, suppose I write a paper on element formation in stars. I cite important papers on the subject including the Caughlin, Fowler, Zimmerman paper and include references to specific paragraphs, suggesting that if the other work is handy to the reader, they can JMP over to that other paragraph and read it for more information. Just because I say something like "See page 3, second paragraph" doesn't make my work derivative of the other. My map is not one of non-literal copyrighted elements (as might be arguably the case in a book like "The Wind Done Gone."
Ok, so suppose instead of being published in a journal, the paper is instead published in an astrophysics anthology along with several of the papers I cite. Again, one book with one cover. And infact the publisher probably owns the copyrights for the ordering and arrangement. But again, that is not a derivative work.
The real question is whether "based on" in the GPL has the same meaning as it does in the Copyright Act (where it is *clearly* defined in such a way that linking would not qualify) and whether "aggregation" means the same thing as preparing a collected or compiled work. If that is the case, then the GPL does not control linking and you can create a wonderful proprietary application that .
Note, however, that a lot of other issues could create derivative works in software. For example, reusing graphical design aspects, or allowing other non-literal creative elements of a program to creap into the new work would imply derivation.
Again IANAL, but this is my $0.02
Then please take this as 2 feature requests:
1) Provide an option for administrators to disable any non-strict-mode settings. I.e. have a super-strict mode that clients cannot get out of.
2) Provide a syntax in table creation to cause the CREATE TABLE to error out if the requested table type is not available.
Both of the situations I described above have to do less with lack of options in MySQL, but rather the lack of central control either from an application developer perspective (no way to *force* InnoDB tables or else not create tables at all) or from a DBI perspective (no way to *force* real data bounds checking). Hence the "options" you speak of seem like they should be *requirements* rather than options for real applications.
There is one issue here.
Suppose we are collecting information. We want to collect information as to a primary place of residence for individuals on a survey but they are allowed to decline to provide that information. When they provide the information, we populate the address tables (normalizing addresses is a *real* pain) and provide a link to the address in the individual table in the primary_residence field.
When primary_residence is NULL we *know* that this means that it is unknown. Our data model expects that for *every* respondant, there is *exactly* one primary residence. Hence enforcing this functional dependency makes a great deal of sense in any normal form.
In a similar case, suppose we store expiration dates for tax rates. NULL means "no plans to change the rate yet" in other words "will probably change at some unknown time in the future." NULL is probably more semantically correct than an infinity timestamp or other fun kludges.
In short, I see nothing in BCNF, 4NF, or 5NF definitions which excludes NULLs used in this manner (i.e. information which is presumed to exist but is unknown at the moment). If the information exists, even if it is not known, the database can and should be designed to assume that it does exist.
The other issue however is that sometimes NULL is use to represent data which does not exist.
For example, consider:
create table employee (
person_id int primary key references person(id),
hourly_wage numeric,
monthy_salary numeric
CHECK hourly_wage is not null OR monthly_salary is not null,
CHECK hourly_wage is null OR monthly_salary is null
);
In this case, we know that either hourly_wage or monthly_salary will be null for every row but not both.
This is dangerous for a number of reasons:
1) Yes, it breaks 3NF because NULLs representing "information does not apply" suggest that there is no functional dependency on that record. In the above example, it would be better redesign the db to avoid those NULLs.
2) It introduces semantic ambiguity into the relational model because elsewhere NULL may be used to mean "UNKNOWN."
Hope that helps.
So how do I ensure that only transactional data storage engines are used when I distribute my accounting application?
Oh wait, I can't because if you have people set up the server wrong it will silently drop my requests for transactional storage....
How do I ensure that no client apps disable strict model and start posting transactions for 2007-02-30?
Oh wait, I can't because any client can turn strict mode off....
Better stick to PostgreSQL for areas where transactional support or data integrity matters.....
Given that normalization definitions deal with data dependencies, I dont see how this is a normalization question.
It is a valid database design question but I would be inclined to use NULL in that table unless I expected at least one person not to have a valid SSN. In which case I would break it off.
SSN is functionally dependant on the person if it exists for the person. However, if the person does *not* have a SSN, then it is not functionally dependant for every portion of the set.
However, nothing prevents you in PostgreSQL from defining a pseudo-Primary Key with NULLS. Something like:
CREATE UNIQUE INDEX foo ON bar(baz) WHERE abc IS NULL;
In that case, X, NULL can only occur once in a table. That can be useful behavior too, for example when you have a table like:
CREATE TABLE tax (
coa_id INT NOT NULL REFERENCES coa(id),
rate NUMERIC NOT NULL,
expires DATE,
);
CREATE UNIQUE INDEX foo ON tax(coa_id, expires);
CREATE UNIQUE INDEX bar ON tax(coa_id) WHERE expires IS NULL;
This allows you to have one rate for each account which is expected to be valid until some unknown point in the future.
You mean like XML? Does MySQL have native xpath capabilities now?
Or do you mean something else?
In general I have generally found that relational storage of *all* information is generally preferred...
1) MySQL is a lot worse than SQLite for a lot of things.
2) Mybe Sun can make MySQL into something not as sucky as it is today?
Domestic and Economic areas:
I: Economic:
A: Renew tax credits for alternative energy development.
B: Create new form of tax credit for energy conservation and distributed development (methane digestors at dairy farms and the like)
C: Work with industry on reducing scope of patents in IT-related fields.
D: Push for reform of copyright law.
II: Criminal Justice
A: Create a grading system for prisons based on recitivism rates.
B: Create a task force to study effects of sentensing lengths on recitivism rates.
C: Work with states to lower recitivism rates by adopting optimal sentensing lengths based on above studies.
III: Domestic Security and Civil Liberties
A: Create open public forums on issues relating to domestic security. Open up general discussions on abstract security issues relating to aircraft, airports, infrastructure, and federal buildings to the public.
B: Undertake a complete survey of hydroelectric dams in this country for vulnerabilities to attack. Scenarios would include small and large amounts fo conventional low and high explosives, and also thermobaric attack profiles.
C: Undertake a study of the impact on civil liberties of various post-9/11 measures.
D: Work to create a rationed guest worker program to undermine illegal immigration.
E: Work on a controlled legalization of Marijuana to undermine drug cartels which have been ammassing enough firepower to challenge the Mexican Army.
1) Import/export of Cannibis would be prohibited, except as allowed by treaty.
F: Work towards a streamlined procedure for deporting illegal immigrants with dependant American citizen children. The children would be given US passports in the process of deportation. The passports might be applied for and then distributed through local consulates, or might be distributed prior to deportation.
Foreign Policy:
I: Iraq
A: Make it *clear* that our continued involvement in Iraq requires certain rules enforced by the Iraqi government including no official or unofficial use of any militia units by any government agencies.
B: Revoke all contracts with external security firms in Iraq until further notice.
C: Make it *clear* that our continued involvement in Iraq will not exceed our welcome by the duly elected Iraqi government.
D: Seek help from The Arab League and the European Union for peacekeeping operations in Iraq under EU and AL flags.
II: Afghanistan
A: Undertake a review of Afghanistan's Constitution and its impact on governability of the country. Provide advice on this matter to that government in the form of a declassified report.
B: Request permission from Pakistan's government to pursue insurgants across the border.
III: Israel/Iran/Syria
A: Continue Bush's policy of delaying/obstructing delivery of aid when Israel is not playing well with the Palestinians.
B: Continue pushing for a Palestinian State.
C: Declare that we stand for defending the Green Line in Israel. We have no position on attacks against IDF targets outside the Green Line.
D: State to Iran and Syria that we are prepared to give a security guarantee provided that:
1) all nuclear weapons programs are verifiably halted
2) all enrichment for fissile materials is halted
3) they help a stable Iraqi government emerge
If you define a terrorist as an individual or organization which attempts to inspire terror as a way of pushing through some political agenda, that would suggest that the US Gov't is more successful in this area than Al Qaeda....
On the other hand, I would think that health insurance companies might like very much to be able to use racial data for determining premiums. After all, if blacks are more likely to have poverty-related diseases, then even wealthy blacks ought to have higher health insurance rates, right?
Well, the issue is that early grain uses (which involved sprouting the grain) made it susceptible to fermentation. The reason is that the sprouting process makes the seed produce amylase which in turn converts starches into sugars.
I am not sure at all that alcoholic beverages were first created by holy men. More likely they were created by accident at first and quickly deified. Mead, Wine, and Beer all have been revered as sacred in ancient cultures.
I suppose you are right-- RMS wants people to use his software. If he releases a GPL v3.1 which removes the system libraries exception, then all of GNU becomes tied to HURD's release, sorta like Duke Nukem Forever* ;-)
*I have a long-running joke that Duke Nukem Forever has been repeatedly delayed due to the lack of a suitable HURD release to play it on....
Could happen.
Remember Thomas Bushnell? For all the ideas that free software == free speech, Thomas Bushnell was required by Stallman to *resign* from the position of being primary HURD architect (and the project has certainly not recovered from that). Why was he asked to resign? Because he (publically) argued that the GNU Free Documentation License was not free enough to be universally consistant with Debian's Free Software Guidelines.
Is there a model in place to handle the centralized liability that this would create? I mean, right now you pay for car insurance because you control the vehicle. If the contol occurs from a computer made by GE, then if a fatal crash occurs, they might be liable. The technology is mature enough we could have self-driving cars today (heck, we could have had them nearly a decade ago!) but I suspect that liability concerns are a big part of the reason we don't.
Even if self-driving cars are safer than human-driven ones, then it seems to me that it might still be prohibitiv to buy insurance.
It is probably better to say that it is a good thing to review new technology for potential advantages and adopt it where it makes sense. Of all the strategies I can think of nothing is more ineffective than adopting technologies because they are new.
First, I don't dispute your main points. Outsourcing generally does add to the response time and cause difficulties in business process management even beyond what you have listed. However, targetted outsourcing can off-set some of these (and only sometimes) by bringing in experts in certain types of business management. In short in the *right* places it can increase rather than decrease available IT resources.
Now, on the balance, I think that outsourcing today is *generally* done with the wrong parts of IT and generally done with the wrong goals and hence creates far more problems than it solves, but I don't think this is inherent in the idea of outsourcing some things outside areas that a business is good at.
First, IT is not a strategic investment anymore in terms of big business decisions. It is an operational investment. In short, the technology and adoption has improved to the point that one has to invest in it for operational reasons but that these don't create strategic considerations.
Secondly, the IT department isn't going to die just because Google now offers office apps. If you want to run any non-trivial computing and communications infrastructure, you need an IT department. Databases of customer data? Accounting data? etc? Sure some of this can be outsourced, but how far do you want to take it? You still want to have some control over things like authentication services, etc.
The way I see it, the shift to SAAS simply broadens the choices for career development. It doesn't eliminate corporate IT departments but simply offers them some competition.
I have been reading a fair bit of legal analysis (IANAL) relating to the GPL v2 and have been discussing various ambiguities relating to the GPL v3 with people at the SFLC. These licenses *do* have some ambiguities (though I think they are less of an issue for the GPL v2).
The major issue for the GPL v2 is that it is not 100% clear where the boundary relating to mere aggregation is. In general it is easy to read "a work based on the original work" meaning derivative work (i.e. a transformation or adaption of the original work in the same way that a movie may be based on a book, or a sequel may be based on another book), while aggregation seems to read as a collected or compiled work, but these simple interpretations are at odds with the FSF's interpretations. I.e. dynamic or even static linking would seem to create (possibly non-literal) compilations under copyright law, not derivations even if the linker strips out unused portions (this is because that process would not be creative enough to create a *new* copyrighted work in the form of the new library code). Hence the simple reading of the GPL v2 would seem to allow one to link proprietary applications to, say, GNU Readline. This question has not been resolved in court yet.
The GPL v3 has the same issue, but adds a few more. For example, does section 7, paragraph 2 govern sections of BSD code included verbatim in a GPL v3 application? I.e. must one be allowed to change the license of a file to the GPL v3 in order to call it compatible? (Eben Moglen says "Yes" while Richard Fontana says "No"-- both are members of the SFLC and both were involved in the GPL v3 development process.)
There are also a few false ambiguities-- for example the question as to whether mere use of software inside an organization might ever one to license patents out (the relevant section of the GPL v3 only applies to explicit patent licenses), though clearly one would want to stop using software before filing patent suits due to patent retaliation clauses.
It seems to me that there is a fundamental logical problem in the analysis.
IMO, the scientific method is all about information processing. All scientists do is derive experiments to gather information and process that information. If information processing is not capable of arriving at an experimental conclusion, then this generally means that one is either processing insufficient data, invalid data, or processing valid data wrong. Either way, it means that the information processing regimen is insufficient for the science.
Now, since we assume the information processing is flawed or insufficient in a mismatch, then we have to understand that holes in information processing only tell us that our existing conclusions are wrong. Hence this is *not* falsifiable and is no different than intelligent design.
Best Wishes,
Chris Travers
There is always benefit to having multiple paths available. Separate tools generally excel at separate tasks.
Just because H.323 was available, does this mean that the IETF should have adopted it instead of RTP/SIP? Obviously when one is largely dealing with TCP/IP networks only, the latter standard is *far* better, but H.323 can be used there too (as well as many other environments-- I would expect it should be reasonably simple to run H.323 over a cell-switched or circuit-switched network for example, without the complexity that it introduces on a packet-switched network.
My first response on hearing about the deal is "no good can come of this." The fact is-- these companies directly compete in most areas and one is *far* stronger than the other.
Apple was an unusual case because Microsoft was investing in them during the hight of the DoJ investigations of their anticompetitive behavior. My own thinking is that Microsoft was terribly afraid that Apple exiting the market would mean that they would be broken up or crippled behind very onerous consent decrees. Apple thus was actually the *stronger* partner in that negotiation because they had something that Microsoft wanted even more than they needed continued support. With Novell, it is not quite the same case.
I dunno. OfficeOpenXML vs ODF is a fairly complex and technical debate which I have generally been loath to jump into. In general, there are things I like about both standards. Programatically OOXML is a better standard in terms of transformation while ODF is a better standard in terms of automatic generation. In general, I think that OOXML is a better standard for office suites' internal formats, while ODF is a more open standard for connecting to non-XML formats. Also the patent issue of OOXML is a major cause of concern.
The problem is that people generally think that one of the two must succeed at the others' expense. Personally I would like nothing better than for both sides to be willing to push for universal open source adoption of both formats without fear of patent infringement suits.