Ladies and gentlemen of the school board, let's play that ever-fun game, Set Your Priorities!
And what are our choices for this year? Where should we be spending our time and money? Pick carefully, the ones you want will be included in the curriculum and the ones you don't want won't be taught!
Science
Library
Music
Physical well-being
How to be good copyright-following citizens
How your corporate sponsor creates a great product
This demonstrates the difference between commercial/professionally run products and what can be a very ad hoc management style for open products.
A commercial organization receives a DR and reviews it. The DR is assigned a priority and a severity. Being obscure and performance related, I'd guess that it scored low on both. It doesn't impact security, it doesn't rear its ugly head often. So it won't impact many users, and presumably, the impact won't be that great. As such, and assuming that you have limited resources devoted to a product, it doesn't exactly float to the top of the heap.
But from the standpoint of code, the defect *might* be interesting! And in a looser environment, interesting trumps utility. Also, the impacted source might be more isolated... meaning to the volunteer "dive right in" developer, it is a more attractive problem to handle.
I'm not trying to defend Oracle or condemn the MariaDB team. I'm using this as an example of how different development processes and practices (highly managed/cathedral vs. open-uncommitted/bazaar) might yield different results. And how different group goals (further integration of MySQL into the Oracle family vs. ??? for MariaDB) might impact where efforts are place.
Yet a smart calculator can get past the "here's the formula for this" phase to "here's a way to think about and experiment with data" phase. After all, do you know how to compute square roots by hand? I learned it in fifth grade (some long time ago), and haven't used it since, well, fifth grade.
Consider my daughter (now in college) studying biology. I'd help/watch her with homework over time while she was in high school. TI-84, using statistical software. Having the tool, and quickly being able to go from raw data to processed data? Priceless, in that she could experiment. She could try different sets and learn the impact of source changes on the resultant. I wouldn't have wanted her to have had to calculate that by hand. She'd have learned technique, but not comprehension.
And I'm happy with her there. I'd rather she know how to work with data and build from there. That's the future.
TLDs are a thing of the past, or will be. The TLD explosion will hasten that.
Remember the early days of ebay? How you could peruse ALL of the new postings for a day in "computer hardware" (one single category) in ten minutes? Yes, you would go to computer>hardware to get to the category, and that's what you did.
Now ebay has been overrun by online stores and bulk postings, a single ID posting hundreds or more items per day. A virtual online catalog for thousands of sale-by-the-shovel retailers. And everything is found via search. (Hey, try posting computer hardware in "dolls". People will find it and it'll sell for the same price you'd get elsewhere.)
This is the future of domain names. You search by company name in the address field, it goes where you want to, your browser/search engine learn your preferences. In five years, no one will care if you are prostitutes-international.xxx or im.a.who.re !
Yeah, I anayzed the CORDICS. And the source implementation of atan2, etc. And optimized for cos = 1-sqrt(sin). (By the way, sqrt is expensive, too). And looked at integer approximations.
The key was I ONLY NEEDED ACCURACY WITHIN 5%!!!!! As others have pointed out, understanding the problem speeds things along nicely. Reviewing the source code showed that the lib calls used hardware doubles. Cycle-expensive. Modern processors, FP and INT are pretty close when compared to trig functions. And only needing very low precision meant the taylor expansions were around two mults and an add. Three and two for some of the more expensive ops, but the trig ran more expensive there as well.
Side note: Going to int was considered as well, but adding in the renormalizations added time, too. And I usually avoid div when I can.
Side note: The power of 1.5 can be done as a cube and a square root (sqrt(x*x*x)). It is cheaper to do x^1.5 as x*sqrt(x), saving a couple of mults...
Did the math... around 100K trig calls per frame. Maybe a few thousand points, but each calculated the intersection of multiple phased waves. And my code was faster than a table lookup!
Bullshit. You want to develop game engines, you have to know how to move things around. You have to know how to create a world. You have to understand what your team is doing, and understand it at a gut level. This isn't rendering, this is creating a world.
So learn your math well. It will give you an additional layer of depth that will differentiate you from those who don't.
Disclaimer: I just coded up some low-precision trig operations. I'm generating some simulated waves in the world I'm creating. Costing hundreds of trig calls per frame, and figure 50 frames per second... My routines are around 3 to 5 times faster than system calls, you want to look up that hundreds per frame at 50 fps again? It means I have a lot more CPU available for other tasks. Math. Taylor series. Error analysis. Makes me the person that gets hired instead of the one that gets passed over.
11 environments (ObjC, Mac, C#/WPF....NET) over five years, so I think he means around 5 months apiece! In other words, a beginner/dabbler who hasn't been involved in a project of any depth...
The news purveyors are complaiining that a summary of the article is being presented. People only read the summary and don't click to see the whole article. Ad revenue due to the news purveyor is lost.
This seems similar to the original arguments against selling music by the track instead of the entire CD. The "old model" was that the purchase package was a full CD (with a few good songs and a lot of dogs). This parallels showing a whole page (with a few interesting paragraphs and a lot of filler. The content owners wanted to sell the whole package, not just the highlights.
The new model is letting the listener hear a short clip (the paragraph on the aggregator's page), and then buying an entire song (viewing the whole article on the host page) if interested. Selling a whole CD (buying the magazine/newspaper or hopping to linked articles at the host's site) may be done if there's sufficient "good" content. And once on the host site, the viewer may well view more than the one article.
This seems to work well for the music industry. Yes, the model has changed. Yes, they have adapted. The print world needs to examine this model, use it, adapt.
Understanding why the math works makes the programs work. Understanding probability and statistics make my inline sampling calculations correct. Understaning how spline calculations work make my curve approximation code (or even the use of curve approximation libraries) correct.
Yes, there are a lot of good libraries out there. They are optimized. They are error-correcting. They are correct. And knowing what they do and how they work enables you to use them effectively.
When you talk to your clients (or your bosses) and they ask you about how you did something, the ability to pull the core math and explain it will go very far.
It is kind of like lifting weights. The lifting isn't its own end. It makes the daily (carrying 40 pound boxes of cat litter into the house) mundane. So it is with math. Understanding simple things as polynomial interpolations for higher-order polynomials can make or break your ability to project storage estimates. Understanding O() notation will help you program well.
Don't scrimp on the math. There are enough bad systems out there for other reasons already.
Okay, you want to know how many ways the world can end? Pick up a collection of Ballard. Drowning. Drought. Giant crystals. On and on. Each the despair and hopeless of mankind at its final moments.
Funny, this thread reads like "the great writers of the previous generations..."
Curious about the SDM32F4 Discovery -- Can you point out a reference to a good open source tool chain, as well as a supplier for the base board? The st.com web site won't let me order one, other places are out of stock...
We moved to a mix of NAS and Cloud a few years ago. We wouldn't go back.
No more time spent on our servers. No more worrying about patches, upgrades, hardware failure, etc. No staff time lost to systems maintenace, backups.
We use the cloud for most project storage. Always [sic] available, at the office, at home, at client sites. Added benefits include systems backups, syncing folders, etc. Requires some trust in the vendor maintaining system integrity, but the risk is lower than at-office implementation.
We use two NAS devices for corporate data and archival storage. Each has a mirror set; I just like two devices as well because (a) they are dirt cheap, and (b) it gives me some level of redundancy in case one of the boxes goes and dies. They sync to each other. We periodically burn DVDs for offisite backup, multiple copies. This is the only weak point; maybe at some point in the future I'll add a third, off-site, sync'ed system.
In all, it works well. IMHO, it provides more than a single onsite server would provide, at a lower cost.
Software patents work without source code "work" (please note that I'm using quotes to denote the process, not the validity of the process) because the patent discloses the technique.
Having the source code in a particular language is irrelevant. The source code is not the invention. The method behind the source code is the invention.
Beside, what relationship does the source code have with the invention? I'll postulate: None. First, the source code is an intermediary between the idea and the execution process. Any of a number of intermediaries can be used. Should revealing the source code in C++ mean that a parallel implementation in Fortran is allowed/does not violate the patent? Second, even using the same source code, what is the impact of compiling to a different architecture?
No, source code has no value except as A METHOD of explaining the idea. It is not the idea.
And so many of these products are still extremely effective. Some, when you ignore biased studies, still more effective than the newer, patented meds that replace them.
Pharma pushes new meds over old ones because they are patented and more lucrative, not because they are better. (Yeah, a gross generalization, but true in many cases.) Pharma takes old meds about to expire, performs a minor tweak or a new study, and gets a new patent/patent extension on what is essentially the same medicine. No new development, just efficacity studies on the new use. I know someone who took viagra (effectively) (patented in 1996) for pulminary hypertension. $1500 a month (thank you big pharma). By the way, ED Viagra was due to expire in 2012, but with the "new" pulminary hypertension effect, the patent now expires in 2020. Who knows what use they'll find for it in 2019?
Here are some questions: WHY is it so expensive to bring a new drug to market? A lot is currently consumed in paperwork and trials. How effective are these trials? Can they be made more effective? If we reduced the trial set to half of what it is now, how much additional risk would there be? 5%? 10%? From a societal standpoint, would the added risk be offset by the impact of increased availability and improved general health? Would the risk of 100 additional deaths be offset by making a drug more available to 1000 additional people? Should cheaper/faster drugs be made available if people agree to limited liability on lawsuits (the Vioxx effect)?
Hey, I still have my Macintosh SE/30 and it still runs. Surely we can use it to write a computer virus that will wipe their databases, rewrite their laws, inject "Earth has Prior Art" all over their WikiPodia, spoil the milk in their fridges, raise their postal rates, and put big X marks at all of the wrong places on images on their porn sites!
Same thing happened when I bought a car. Okay, I'm not a car expert, but I really wanted a Camry. Went to a Chevy dealer, they both begin with C and that confused me. Doors, windows, wheels, I could have SWORN I got a Camry. Boy, was I surprised when I got home and finally saw the GM logo on the key!
I'm of the mindset that even a little alcohol in your system should keep you off the road. Alcohol affects people in different ways, what may be fine for you isn't fine for me. If you wan to have a drink, don't plan on driving.
Then please don't drink any orange juice, which can get to up to 0.4% alcohol (around 1 proof).
I'd be tempted to create a Judge Hall of Fame for judges that just plain "get it." For ones who have an understanding of the philosophy and content of the fields upon which they are judging. And produce decisions based upon sound reason.
Judge Posner.
Judge John Jones (Dover: Intelligent Design, everyone should read this at least once a year.
Comments requested: who else should be on this list?
If thier intent is to make a political statement, they will succeed.
If this is a business strategy, it will fail the same way that U.S. health care (ACA) will fail if you require companies to take all subscribers but do not require all people to subscribe.
As a business strategy, they are spreading the tax across IE7 users, a population that is not required to use thier site. IE7 users may choose to go elsewhere ("being insulted" and "higher costs"), which means the *fixed* cost of the support (web site maintenance) is spread across a smaller number of users. From here, basic economics: the fixed cost results in a higher per-user tax. Resulting in fewer users. Cycling to a higher tax again. (This will be true regardless of whether or not some IE7 users upgrade to use their site -- as long as some IE7 users go away, they have a reduced user population and a higher IE7 tax.
The end result will be no IE7 users and fewer users in general.
So they might as well jump to the endpoint: Don't bother coding for IE7 (saving cost), don't tax users (since they aren't using the tax to fund IE7 support), and as long as the drop in revenue/profits is less than the drop in cost, the strategy is successful. A simple log review will give them an estimate of IE7 usage on thier site. This should drive their decision.
Slashdot Mirror
And what are our choices for this year? Where should we be spending our time and money? Pick carefully, the ones you want will be included in the curriculum and the ones you don't want won't be taught!
And the winner is....
From my earlier years: Fast and cheap file transfer? A grad student with a mag tape.
This demonstrates the difference between commercial/professionally run products and what can be a very ad hoc management style for open products.
A commercial organization receives a DR and reviews it. The DR is assigned a priority and a severity. Being obscure and performance related, I'd guess that it scored low on both. It doesn't impact security, it doesn't rear its ugly head often. So it won't impact many users, and presumably, the impact won't be that great. As such, and assuming that you have limited resources devoted to a product, it doesn't exactly float to the top of the heap.
But from the standpoint of code, the defect *might* be interesting! And in a looser environment, interesting trumps utility. Also, the impacted source might be more isolated... meaning to the volunteer "dive right in" developer, it is a more attractive problem to handle.
I'm not trying to defend Oracle or condemn the MariaDB team. I'm using this as an example of how different development processes and practices (highly managed/cathedral vs. open-uncommitted/bazaar) might yield different results. And how different group goals (further integration of MySQL into the Oracle family vs. ??? for MariaDB) might impact where efforts are place.
Yet a smart calculator can get past the "here's the formula for this" phase to "here's a way to think about and experiment with data" phase. After all, do you know how to compute square roots by hand? I learned it in fifth grade (some long time ago), and haven't used it since, well, fifth grade.
Consider my daughter (now in college) studying biology. I'd help/watch her with homework over time while she was in high school. TI-84, using statistical software. Having the tool, and quickly being able to go from raw data to processed data? Priceless, in that she could experiment. She could try different sets and learn the impact of source changes on the resultant. I wouldn't have wanted her to have had to calculate that by hand. She'd have learned technique, but not comprehension.
And I'm happy with her there. I'd rather she know how to work with data and build from there. That's the future.
TLDs are a thing of the past, or will be. The TLD explosion will hasten that.
Remember the early days of ebay? How you could peruse ALL of the new postings for a day in "computer hardware" (one single category) in ten minutes? Yes, you would go to computer>hardware to get to the category, and that's what you did.
Now ebay has been overrun by online stores and bulk postings, a single ID posting hundreds or more items per day. A virtual online catalog for thousands of sale-by-the-shovel retailers. And everything is found via search. (Hey, try posting computer hardware in "dolls". People will find it and it'll sell for the same price you'd get elsewhere.)
This is the future of domain names. You search by company name in the address field, it goes where you want to, your browser/search engine learn your preferences. In five years, no one will care if you are prostitutes-international.xxx or im.a.who.re !
Today is National Sysadmin Day. It is also National Coffee Milkshake Day.
Coincidence? I think not.
Nope, too slow. Two-term Taylor series.
Yeah, I anayzed the CORDICS. And the source implementation of atan2, etc. And optimized for cos = 1-sqrt(sin). (By the way, sqrt is expensive, too). And looked at integer approximations.
The key was I ONLY NEEDED ACCURACY WITHIN 5%!!!!! As others have pointed out, understanding the problem speeds things along nicely. Reviewing the source code showed that the lib calls used hardware doubles. Cycle-expensive. Modern processors, FP and INT are pretty close when compared to trig functions. And only needing very low precision meant the taylor expansions were around two mults and an add. Three and two for some of the more expensive ops, but the trig ran more expensive there as well.
Side note: Going to int was considered as well, but adding in the renormalizations added time, too. And I usually avoid div when I can.
Side note: The power of 1.5 can be done as a cube and a square root (sqrt(x*x*x)). It is cheaper to do x^1.5 as x*sqrt(x), saving a couple of mults...
Did the math... around 100K trig calls per frame. Maybe a few thousand points, but each calculated the intersection of multiple phased waves. And my code was faster than a table lookup!
Look it up, program it, forget it, and leave it for someone else to fix the bugs because you didn't understand it...
Bullshit. You want to develop game engines, you have to know how to move things around. You have to know how to create a world. You have to understand what your team is doing, and understand it at a gut level. This isn't rendering, this is creating a world.
So learn your math well. It will give you an additional layer of depth that will differentiate you from those who don't.
Disclaimer: I just coded up some low-precision trig operations. I'm generating some simulated waves in the world I'm creating. Costing hundreds of trig calls per frame, and figure 50 frames per second... My routines are around 3 to 5 times faster than system calls, you want to look up that hundreds per frame at 50 fps again? It means I have a lot more CPU available for other tasks. Math. Taylor series. Error analysis. Makes me the person that gets hired instead of the one that gets passed over.
11 environments (ObjC, Mac, C#/WPF... .NET) over five years, so I think he means around 5 months apiece! In other words, a beginner/dabbler who hasn't been involved in a project of any depth...
The news purveyors are complaiining that a summary of the article is being presented. People only read the summary and don't click to see the whole article. Ad revenue due to the news purveyor is lost.
This seems similar to the original arguments against selling music by the track instead of the entire CD. The "old model" was that the purchase package was a full CD (with a few good songs and a lot of dogs). This parallels showing a whole page (with a few interesting paragraphs and a lot of filler. The content owners wanted to sell the whole package, not just the highlights.
The new model is letting the listener hear a short clip (the paragraph on the aggregator's page), and then buying an entire song (viewing the whole article on the host page) if interested. Selling a whole CD (buying the magazine/newspaper or hopping to linked articles at the host's site) may be done if there's sufficient "good" content. And once on the host site, the viewer may well view more than the one article.
This seems to work well for the music industry. Yes, the model has changed. Yes, they have adapted. The print world needs to examine this model, use it, adapt.
Turn Left! Or so the Nav system of my unlocked phone said, even though I was in the middle of the bridge...
Understanding why the math works makes the programs work. Understanding probability and statistics make my inline sampling calculations correct. Understaning how spline calculations work make my curve approximation code (or even the use of curve approximation libraries) correct.
Yes, there are a lot of good libraries out there. They are optimized. They are error-correcting. They are correct. And knowing what they do and how they work enables you to use them effectively.
When you talk to your clients (or your bosses) and they ask you about how you did something, the ability to pull the core math and explain it will go very far.
It is kind of like lifting weights. The lifting isn't its own end. It makes the daily (carrying 40 pound boxes of cat litter into the house) mundane. So it is with math. Understanding simple things as polynomial interpolations for higher-order polynomials can make or break your ability to project storage estimates. Understanding O() notation will help you program well.
Don't scrimp on the math. There are enough bad systems out there for other reasons already.
Okay, you want to know how many ways the world can end? Pick up a collection of Ballard. Drowning. Drought. Giant crystals. On and on. Each the despair and hopeless of mankind at its final moments.
Funny, this thread reads like "the great writers of the previous generations..."
Curious about the SDM32F4 Discovery -- Can you point out a reference to a good open source tool chain, as well as a supplier for the base board? The st.com web site won't let me order one, other places are out of stock...
We moved to a mix of NAS and Cloud a few years ago. We wouldn't go back.
No more time spent on our servers. No more worrying about patches, upgrades, hardware failure, etc. No staff time lost to systems maintenace, backups.
We use the cloud for most project storage. Always [sic] available, at the office, at home, at client sites. Added benefits include systems backups, syncing folders, etc. Requires some trust in the vendor maintaining system integrity, but the risk is lower than at-office implementation.
We use two NAS devices for corporate data and archival storage. Each has a mirror set; I just like two devices as well because (a) they are dirt cheap, and (b) it gives me some level of redundancy in case one of the boxes goes and dies. They sync to each other. We periodically burn DVDs for offisite backup, multiple copies. This is the only weak point; maybe at some point in the future I'll add a third, off-site, sync'ed system.
In all, it works well. IMHO, it provides more than a single onsite server would provide, at a lower cost.
Software patents work without source code "work" (please note that I'm using quotes to denote the process, not the validity of the process) because the patent discloses the technique. Having the source code in a particular language is irrelevant. The source code is not the invention. The method behind the source code is the invention. Beside, what relationship does the source code have with the invention? I'll postulate: None. First, the source code is an intermediary between the idea and the execution process. Any of a number of intermediaries can be used. Should revealing the source code in C++ mean that a parallel implementation in Fortran is allowed/does not violate the patent? Second, even using the same source code, what is the impact of compiling to a different architecture? No, source code has no value except as A METHOD of explaining the idea. It is not the idea.
And so many of these products are still extremely effective. Some, when you ignore biased studies, still more effective than the newer, patented meds that replace them.
Pharma pushes new meds over old ones because they are patented and more lucrative, not because they are better. (Yeah, a gross generalization, but true in many cases.) Pharma takes old meds about to expire, performs a minor tweak or a new study, and gets a new patent/patent extension on what is essentially the same medicine. No new development, just efficacity studies on the new use. I know someone who took viagra (effectively) (patented in 1996) for pulminary hypertension. $1500 a month (thank you big pharma). By the way, ED Viagra was due to expire in 2012, but with the "new" pulminary hypertension effect, the patent now expires in 2020. Who knows what use they'll find for it in 2019?
Here are some questions: WHY is it so expensive to bring a new drug to market? A lot is currently consumed in paperwork and trials. How effective are these trials? Can they be made more effective? If we reduced the trial set to half of what it is now, how much additional risk would there be? 5%? 10%? From a societal standpoint, would the added risk be offset by the impact of increased availability and improved general health? Would the risk of 100 additional deaths be offset by making a drug more available to 1000 additional people? Should cheaper/faster drugs be made available if people agree to limited liability on lawsuits (the Vioxx effect)?
Hey, I still have my Macintosh SE/30 and it still runs. Surely we can use it to write a computer virus that will wipe their databases, rewrite their laws, inject "Earth has Prior Art" all over their WikiPodia, spoil the milk in their fridges, raise their postal rates, and put big X marks at all of the wrong places on images on their porn sites!
Same thing happened when I bought a car. Okay, I'm not a car expert, but I really wanted a Camry. Went to a Chevy dealer, they both begin with C and that confused me. Doors, windows, wheels, I could have SWORN I got a Camry. Boy, was I surprised when I got home and finally saw the GM logo on the key!
I'm of the mindset that even a little alcohol in your system should keep you off the road. Alcohol affects people in different ways, what may be fine for you isn't fine for me. If you wan to have a drink, don't plan on driving.
Then please don't drink any orange juice, which can get to up to 0.4% alcohol (around 1 proof).
I'd be tempted to create a Judge Hall of Fame for judges that just plain "get it." For ones who have an understanding of the philosophy and content of the fields upon which they are judging. And produce decisions based upon sound reason.
Judge Posner.
Judge John Jones (Dover: Intelligent Design, everyone should read this at least once a year.
Comments requested: who else should be on this list?
If thier intent is to make a political statement, they will succeed.
If this is a business strategy, it will fail the same way that U.S. health care (ACA) will fail if you require companies to take all subscribers but do not require all people to subscribe.
As a business strategy, they are spreading the tax across IE7 users, a population that is not required to use thier site. IE7 users may choose to go elsewhere ("being insulted" and "higher costs"), which means the *fixed* cost of the support (web site maintenance) is spread across a smaller number of users. From here, basic economics: the fixed cost results in a higher per-user tax. Resulting in fewer users. Cycling to a higher tax again. (This will be true regardless of whether or not some IE7 users upgrade to use their site -- as long as some IE7 users go away, they have a reduced user population and a higher IE7 tax.
The end result will be no IE7 users and fewer users in general.
So they might as well jump to the endpoint: Don't bother coding for IE7 (saving cost), don't tax users (since they aren't using the tax to fund IE7 support), and as long as the drop in revenue/profits is less than the drop in cost, the strategy is successful. A simple log review will give them an estimate of IE7 usage on thier site. This should drive their decision.