The facts from their own chart: 10 billion in physical game sales vs 200 million in app sales. Even if it increases by 200 million a year, it will take a long time to catch up.
Not $200 million/year, $200 million per month. Compared to 840 million/month ($10 billion/year) for the physical game market.
Of course, people with non iOS game hardware will keep buying games, and the App store will expand the market. But it seems to be a new force.
While Ioannidis, the author of the original work that was discussed here may be correct that a large fraction of a very specific type of clinical research findings, are incorrect, there is no reason to believe (from his published work) that "most published research findings are false". Most of the ones he looked at were not reproduced, but they had well understood limitations. My papers, I can assure you, are only incorrect about 10% of the time.
If you compare the tariffs for O2 service (12 month contract) before the iPhone, you could get 200 min and 400 texts for 30 pounds/month http://shop.o2.co.uk/tariffs/12_months or 500 min/100 texts on an 18-month contract. For 5 pounds more, you get unlimited data access with the iPhone. While there may be cheaper ways to get cell phone service in the UK, if you want O2, the rates for the iPhone are not much higher than pre-iPhone.
Putting aside for a moment the question of whether Genetic Association Studies - the focus of the research paper - are representative of "Most Science", the article does not say the analysis is invariably sloppy, it says it is often mistaken. For genetic association studies, this is not surprising, since it is very difficult to publish a negative result. So, small studies that show a statistically significant relationship are published, but small studies showing no relationship are not. Then, when larger studies are done, the small studies that had the "significant" relationship because of a fortunate or unfortunate set of samples is not confirmed. Indeed, this is what the research article points out; if your threshold for statistical significance is 0.05, then you will report that a chance relationship is significant once in 20 experiments. But, if you can't publish the 19 negative experiments, then lots of chance results get published.
But Dr. Ioannidis has a very narrow definition of science - he only includes statistical studies that use p 0.05 as a threshold for significance. There are, of course, lots of papers that do not show p-values - the purification of a protein, the determination of a genome sequence, the identification of a new fundamental particle. In many cases, p-values are not provided because they are not considered informative - something that happens when the p-value is much much much less than 0.05 (I like my p-values less than Avagadro's constant. With that p-valuep, I think most of my results are correct.)
And, of course, the WSJ misses all of this. The point of the research paper is that you can do everything right, and still be mislead with marginal p-values (0.05). Not sloppy, just not significant enough. We could, of course, require more stringent values, but then we would miss the genuinely rare, but important results.
As the research article points out, results that are reproducible are, in fact, quite likely to be correct. It is perhaps useful to distinguish between science as a paper and science as a process. Most results that stand up to scientific scrutiny over a period of years (that any one cares enough about to validate), are (probably) correct. In some disciplines, which rely heavily on modest thresholds for statistical significance, many results cannot be confirmed.
While I am unfamiliar with the arguments that most scientific theories are unfalsifiable, I don't have any problem imagining reasonable tests of evolution that are quite falsifiable (but perhaps I am missing the point).
Part of the problem, of course is that "evolution" means many things to evolutionary theorists, so one needs to be a bit more specific talking about what part of the "theory of evolution" might be tested.
Here is widely agreed upon tenet of evolution that I think is easily falsifiable: "all life on earth evolved from a common ancestor".
There is considerable evidence supporting this hypothesis today - all organisms use essentially the same genetic code; all organisms appear to share at hundreds of proteins with most others, and every organism shares more than half of its proteins with some other organisms, no matter how "weird" the organism is. IF one found an organism (in Nature) that used a different genetic code, or that used a completely novel set of proteins, that organism could not reasonably be said to have evolved from the organisms we know. Since it is hard to argue that alternative genetic codes would be dramatically less efficient than the one that is used, and it seems clear that many organisms use different proteins for common tasks, so that an organism with completely novel proteins is certainly imaginable (there is an enormous number of possible proteins), there seems no inherent reason why organisms with different codes and different proteins might exist. None have been found.
And, I have no doubt that we could DESIGN an organism that used a different genetic code (though I think it would have to be used to encode proteins used in other organisms.
So, I think the hypothesis "all life on earth evolved from a common ancestor" is testable and falsifiable. It will be interesting to see how life on other planets looks.
It's rare to find so much misinformation at Slashdot, and that's saying something.
Humans and chimpanzee DNA are very similar, there are apparently about 40 million differences (out of about 3 billion positions) between chimp and human DNA; in protein coding regions, the number of differences is much smaller.
Humans and mice, on the other hand are far more evolutionarily distant (80 million years since the last common ancestor, compared with 5 million, or less for chimps). In protein coding regions, mouse and human DNA sequences are about 80% identical, on average, but outside protein coding regions, the level of sequence similarity is no higher than would be expected by chance. (This large difference was one of the reasons the mouse genome was sequenced after the human genome - sequences that were more similar than chance were expected to have a function.)
While plants and animals (and bacteria) share a large number of proteins that do similar things, their DNA sequences do not share any significant similarity except in protein coding regions for very highly conserved proteins.
What all of this has to do with unpublished Russian studies on genetically modified plants, I cannot imagine.
While money market funds can lose money (principle) in principle, because they are not insured, I believe the money market associations will point out that no one has EVER lost money in a major money market fund. The types of securities they are allowed to invest in (mostly government notes and commercial with very short terms) are extremely low risk.
In contrast, Savings and Loan accounts, though insured, had considerably higher risk because they were backed by long term mortages. During the Savings and Load crises, many of these mortages were for commercial property that was highly speculative and defaulted.
Although money market rates should track short term interest rates (which are relatively high right now, compared to the past few years), some institutions seem to be able to do much better than others. ING pays much better, right now, than most mutual fund (Fidelity, T Rowe Price) money market funds.
The comment that this study contradicts the results of a previous http://dukemednews.duke.edu/news/article.php?id=50 56 stenting study is incorrect. While that study found a different in the number of adverse effects, the difference was not statistically significant. While the headline for the earlier press release suggested some benefit, the study itself did not.
I had to start prepping for the 360 almost 40 years ago. I learned about punch card machines, and which cards to put at the front of the deck to make certain that the compiler ran, or the program ran. And, of course, I prepped by sleeping a lot during the day, so I could submit jobs in the evening and get them back late at night.
About 25 years ago, I bought a very inexpensive digital watch that was 'glow-in-the-dark'. On the back was a radioactivity symbol that indicated the watch contained 200 mCi of 3H. As a molecular biologist who became very very careful when working with 5 mCi of 32P (a much stronger emitter) or 3H-thymidine, the idea of wearing 200 mCi of 3H seemed quite exciting.
Indeed, I believe there was a superfund site due to 3H contamination from watch manufacturing.
I suspect that the median age of most scientific software code is at least 10 years, and a substantial fraction is 20+ years old. Bioinformatics programs that I wrote 20 years ago are still being used, as are curve fitting programs (very infrequently). The large linear algebra and differential equation packages are OLD. While computers are being used for many new applications (e.g. watching movies), they still do the things they did 20 years ago, faster, and cheaper, but using old code.
There are important similarities between the information processing and transfer in living organisms and mathematical computation, which have been recognized for more than 50 years (see Gunther Stent's "Paradoxes of Progress" for some essays on the nature of genes and biological information transfer as the central dogma was emerging). But there are critical differences as well, which are
often misunderstood.
The fundamental difference between computing in biology and computing with man-made computers is that biological systems were not designed. This has very important implications for the relationship between biology and computer science:
Biological systems are not efficient.
They rarely find optimal solutions; they are simply functional. They reflect evolutionary history and selection, but the selective pressure
may have occurred millions of years in the past and be largely irrelevant today.
So, to answer the questions posed:
Will biology rewrite computing. No. Current digital computers are far faster, more reliable, and more flexible than biolgical systems over short time spans (seconds to millenia). We used to have biological computers do our taxes in the 19th and early 20th century, but we found digital computers far more efficient and accurate.
How will the nature of (scientific) computing change due to biology integration? Biology is not about being optimal, it is about being functional under a wide variety of poorly specified constraints. Current computer science perspectives suffer from a focus on optimality. Optimality is nice, its mathematical, its efficient, but it often answers the wrong question when applied to biological problems. In biology, the shape of solution space is often more important than optimality.
What will be the biggest issue determining biology-integrated computing? I don't know what this means. I'm not certain there will be an integration, see the first answer.
Is this worth pursuing? Of course. Biology offers new perspectives to computer scientists, and computers are central to understanding biology. But they are not very much alike, and pushing the analogies between the two disciplines will probably be less informative than understanding the differences.
I have the same problem, but I have found that one can set up several SMTP servers for the same account, then after one of them fails, the alternative servers are offered. Once the new server is selected, it is used for the remainder of the session.
The ssh solution is much more elegant (and secure), but out-of-the-box, things work pretty well.
Make 2.6.3 usable, never install a dot-oh version
on
Linux Kernel 3.0?
·
· Score: 2, Interesting
Calling the successor of 2.5 version 3.0 ignores the 2.4 (pre.14 or perhaps.16) debacle and the old adage to never deploy a.0 operating system.
While millions of Linux users were apparently happy with the early 2.4 kernels, those of us with heavy CPU large memory needs were appalled when we watched our computers lock up under heavy memory usage. Yes, we thought we had a usable system at 2.4.14, but then came.15, with file system corruption, so.16 was the FIRST usable version for systems with high memory demand. Wouldn't it be great if 2.6.1 was as robust as 2.2, or 2.4.17, at the beginning?
Since we all know better than to deploy a.0 version, 3.0 must be a non-starter.
by ReadbackMonkey - You are thinking of trademark law. The best example is Kleenex. Kleenex made no attempt to protect their trademark and it has become so prolific in society that it has come to mean disposable tissue in the generic sense.
The best example is "aspirin", which was a trademark of the Bayer corporation. They lost it, and now the term is generic.
Kleenex (and Xerox) are still trademarks under the control of their owners.
This makes it even more difficult to explain the pricing on re-mastered CD's of records. I would be happy to replace my 30 year old record collection, but not at $10 - 15/album. The publisher made their costs when the record was released, why not price the CD sensibly?
I would like to answer several questions that were raised in this discussion.
(1) How does a CS person learn biology? I recommend "Recombinant DNA, A short Course", as an accessible (Scientific American style) introduction to the cloning breakthroughs and discoveries that lead to genome science.
(2) How does a CS person learn "Bioinformatcs"? I strongly recommend "Bioinformatics - Sequence and Genome Analysis" by David Mount as an accessible and extremely comprehensive survey of current approaches in Biological Sequence Analysis.
(3) Why do Biologists use Perl? Much of the information Biologists want is on the WWW, and Perl's LWP makes it extremely easy to get it. We don't use Perl for sophisticated text analysis (similarity searching, motif searching, etc) because the algorithms that are appropriate are typically not exact (or even regular expression) matches. But it's difficult to beat Perl for getting stuff off the WWW.
(4) Why do Biologists use Flat files? Several reasons - (a) the most useful information is sequence information, and it can be read much more quickly out of a flatfile (esp. one that is memory mapped) than a DB; (b) flat files solve some versioning problems that DB's make very complex and slow. (c) Most data providers only provide flatfiles. This will change, however, over the next 2 - 3 years, mySQL and postgresQL are moving into biology labs.
It is very exciting that Bioinformatics has high visibility now, and many people with CS background are considering bioinformatics problems. Unfortunately, many of the introductory books on bioinformatics (particularly the O'Reilly books) do not adequately present the substantial foundations of bioinformatics that have been build over the past 15 - 20 years, and some newcomers are mislead into believing there are simple problems looking for a few good programmers. Most of the simple problems have been solved; many of the complicated problems are challenging not because we do not know enough CS, but because we do not know enough biology.
The Silicon Valley article is a bit misleading, and doesn't accurately reflect the range of distribution alternatives being used for Bioinformatics software. It is certainly true that many Universities claim ownership of computer software copyrights, but it is important to appreciate that there many levels at which the implementation of these policies is decided. For example, both the WU-BLAST and the HMMer packages were developed by researchers at Washington U. in St. Louis. WU-BLASTbinaries are available to academics after an appropriate license is signed, and licensed commercially. HMMer i is available under the GPL but a commercial license is also available.
Likewise, the FASTA package, can be freely downloaded by both academic and commerical users, but must be licensed from the U. of Virginia to be redistributed. This has allowed the software to be widely used by researchers and also incorporated into commerical packages.
As a Bioinformatics researcher and software author, my goal is to have my research and software be used as widely as possible. This improves my ability to obtain future external funding, to get my papers cited, etc. etc. Even at universities like Wisconsin and Stanford, which derive enormous sums from IP licensing, these funds are less than 10% the value of NIH and other external funding. Thus, it is not hard to argue that software licensing policies should maximize the likelihood of external funding, and the widest possible distribution (though not necessarily GPL) is likely to have the greatest impact and long term benefit. (Moreover, once software becomes widely used, it is much more valuable commercially.)
Thus, while a university's Vice-President for Research may be interest in IP licensing, a Dept. Chairman may be more interested in faculty success in obtaining external funding, and a broader software distribution.
Before the Bayh-Dole act, the federal government owned the intellectual property produced by grants and contracts, and it was exceptionally bad at providing the patent and other protections necessary to get discoveries into the market place. Drug companies do not develop and market drugs if they cannot control the patent. Likewise, some software companies may not make the marketing and support investments necessary to commercialize a product if the product is available as open source.
Much of this discussion assumes that the "discovery", or IP, has substantial inherent value that can be exploited without additional investments in support, marketing, etc. For specialized fields like bioinformatics, this is simply not true - the program may provide important new capabilities, but it will have little commercial value without appropriate packaging, integration with other biological information, training, and support. The universities do not do these "commercial" tasks, neither does the government. Companies do this sort of thing, but not without some guarantee (e.g. an exclusive license) that limits their risk.
This comment reflects a widely held misconception about how research at public and private universities is funded. Much less than 10% of the research in the physical or natural sciences, or computer science or engineering at major universities (public or private) is funded by the institution. Almost all research is funded by federal agencies (NIH, NSF, DOD), and internal funds for research often come from indirect cost recovery from the federal grants. States and private endowments simply do not pay for research. (At fortunate institutions, they pay for buildings.)
Slashdot Mirror
The facts from their own chart: 10 billion in physical game sales vs 200 million in app sales. Even if it increases by 200 million a year, it will take a long time to catch up.
Not $200 million/year, $200 million per month. Compared to 840 million/month ($10 billion/year) for the physical game market.
Of course, people with non iOS game hardware will keep buying games, and the App store will expand the market. But it seems to be a new force.
While Ioannidis, the author of the original work that was discussed here may be correct that a large fraction of a very specific type of clinical research findings, are incorrect, there is no reason to believe (from his published work) that "most published research findings are false". Most of the ones he looked at were not reproduced, but they had well understood limitations. My papers, I can assure you, are only incorrect about 10% of the time.
If you compare the tariffs for O2 service (12 month contract) before the iPhone, you could get 200 min and 400 texts for 30 pounds/month http://shop.o2.co.uk/tariffs/12_months or 500 min/100 texts on an 18-month contract. For 5 pounds more, you get unlimited data access with the iPhone. While there may be cheaper ways to get cell phone service in the UK, if you want O2, the rates for the iPhone are not much higher than pre-iPhone.
Putting aside for a moment the question of whether Genetic Association Studies - the focus of the research paper - are representative of "Most Science", the article does not say the analysis is invariably sloppy, it says it is often mistaken. For genetic association studies, this is not surprising, since it is very difficult to publish a negative result. So, small studies that show a statistically significant relationship are published, but small studies showing no relationship are not. Then, when larger studies are done, the small studies that had the "significant" relationship because of a fortunate or unfortunate set of samples is not confirmed. Indeed, this is what the research article points out; if your threshold for statistical significance is 0.05, then you will report that a chance relationship is significant once in 20 experiments. But, if you can't publish the 19 negative experiments, then lots of chance results get published.
But Dr. Ioannidis has a very narrow definition of science - he only includes statistical studies that use p 0.05 as a threshold for significance. There are, of course, lots of papers that do not show p-values - the purification of a protein, the determination of a genome sequence, the identification of a new fundamental particle. In many cases, p-values are not provided because they are not considered informative - something that happens when the p-value is much much much less than 0.05 (I like my p-values less than Avagadro's constant. With that p-valuep, I think most of my results are correct.)
And, of course, the WSJ misses all of this. The point of the research paper is that you can do everything right, and still be mislead with marginal p-values (0.05). Not sloppy, just not significant enough. We could, of course, require more stringent values, but then we would miss the genuinely rare, but important results.
As the research article points out, results that are reproducible are, in fact, quite likely to be correct. It is perhaps useful to distinguish between science as a paper and science as a process. Most results that stand up to scientific scrutiny over a period of years (that any one cares enough about to validate), are (probably) correct. In some disciplines, which rely heavily on modest thresholds for statistical significance, many results cannot be confirmed.
While I am unfamiliar with the arguments that most scientific theories are unfalsifiable, I don't have any problem imagining reasonable tests of evolution that are quite falsifiable (but perhaps I am missing the point).
Part of the problem, of course is that "evolution" means many things to evolutionary theorists, so one needs to be a bit more specific talking about what part of the "theory of evolution" might be tested.
Here is widely agreed upon tenet of evolution that I think is easily falsifiable: "all life on earth evolved from a common ancestor".
There is considerable evidence supporting this hypothesis today - all organisms use essentially the same genetic code; all organisms appear to share at hundreds of proteins with most others, and every organism shares more than half of its proteins with some other organisms, no matter how "weird" the organism is. IF one found an organism (in Nature) that used a different genetic code, or that used a completely novel set of proteins, that organism could not reasonably be said to have evolved from the organisms we know. Since it is hard to argue that alternative genetic codes would be dramatically less efficient than the one that is used, and it seems clear that many organisms use different proteins for common tasks, so that an organism with completely novel proteins is certainly imaginable (there is an enormous number of possible proteins), there seems no inherent reason why organisms with different codes and different proteins might exist. None have been found.
And, I have no doubt that we could DESIGN an organism that used a different genetic code (though I think it would have to be used to encode proteins used in other organisms.
So, I think the hypothesis "all life on earth evolved from a common ancestor" is testable and falsifiable. It will be interesting to see how life on other planets looks.
It's rare to find so much misinformation at Slashdot, and that's saying something.
Humans and chimpanzee DNA are very similar, there are apparently about 40 million differences (out of about 3 billion positions) between chimp and human DNA; in protein coding regions, the number of differences is much smaller.
Humans and mice, on the other hand are far more evolutionarily distant (80 million years since the last common ancestor, compared with 5 million, or less for chimps). In protein coding regions, mouse and human DNA sequences are about 80% identical, on average, but outside protein coding regions, the level of sequence similarity is no higher than would be expected by chance. (This large difference was one of the reasons the mouse genome was sequenced after the human genome - sequences that were more similar than chance were expected to have a function.)
While plants and animals (and bacteria) share a large number of proteins that do similar things, their DNA sequences do not share any significant similarity except in protein coding regions for very highly conserved proteins.
What all of this has to do with unpublished Russian studies on genetically modified plants, I cannot imagine.
While money market funds can lose money (principle) in principle, because they are not insured, I believe the money market associations will point out that no one has EVER lost money in a major money market fund. The types of securities they are allowed to invest in (mostly government notes and commercial with very short terms) are extremely low risk.
In contrast, Savings and Loan accounts, though insured, had considerably higher risk because they were backed by long term mortages. During the Savings and Load crises, many of these mortages were for commercial property that was highly speculative and defaulted.
Although money market rates should track short term interest rates (which are relatively high right now, compared to the past few years), some institutions seem to be able to do much better than others. ING pays much better, right now, than most mutual fund (Fidelity, T Rowe Price) money market funds.
The comment that this study contradicts the results of a previous http://dukemednews.duke.edu/news/article.php?id=50 56 stenting study is incorrect. While that study found a different in the number of adverse effects, the difference was not statistically significant. While the headline for the earlier press release suggested some benefit, the study itself did not.
I had to start prepping for the 360 almost 40 years ago. I learned about punch card machines, and which cards to put at the front of the deck to make certain that the compiler ran, or the program ran. And, of course, I prepped by sleeping a lot during the day, so I could submit jobs in the evening and get them back late at night.
So, what's new?
About 25 years ago, I bought a very inexpensive digital watch that was 'glow-in-the-dark'. On the back was a radioactivity symbol that indicated the watch contained 200 mCi of 3H. As a molecular biologist who became very very careful when working with 5 mCi of 32P (a much stronger emitter) or 3H-thymidine, the idea of wearing 200 mCi of 3H seemed quite exciting.
Indeed, I believe there was a superfund site due to 3H contamination from watch manufacturing.
I suspect that the median age of most scientific software code is at least 10 years, and a substantial fraction is 20+ years old. Bioinformatics programs that I wrote 20 years ago are still being used, as are curve fitting programs (very infrequently). The large linear algebra and differential equation packages are OLD. While computers are being used for many new applications (e.g. watching movies), they still do the things they did 20 years ago, faster, and cheaper, but using old code.
There are important similarities between the information processing and transfer in living organisms and mathematical computation, which have been recognized for more than 50 years (see Gunther Stent's "Paradoxes of Progress" for some essays on the nature of genes and biological information transfer as the central dogma was emerging). But there are critical differences as well, which are often misunderstood.
The fundamental difference between computing in biology and computing with man-made computers is that biological systems were not designed. This has very important implications for the relationship between biology and computer science:
So, to answer the questions posed:
The ssh solution is much more elegant (and secure), but out-of-the-box, things work pretty well.
While millions of Linux users were apparently happy with the early 2.4 kernels, those of us with heavy CPU large memory needs were appalled when we watched our computers lock up under heavy memory usage. Yes, we thought we had a usable system at 2.4.14, but then came .15, with file system corruption, so .16 was the FIRST usable version for systems with high memory demand. Wouldn't it be great if 2.6.1 was as robust as 2.2, or 2.4.17, at the beginning?
Since we all know better than to deploy a .0 version, 3.0 must be a non-starter.
You are thinking of trademark law. The best example is Kleenex. Kleenex made no attempt to protect their trademark and it has become so prolific in society that it has come to mean disposable tissue in the generic sense.
The best example is "aspirin", which was a trademark of the Bayer corporation. They lost it, and now the term is generic.
Kleenex (and Xerox) are still trademarks under the control of their owners.
This makes it even more difficult to explain the pricing on re-mastered CD's of records. I would be happy to replace my 30 year old record collection, but not at $10 - 15/album. The publisher made their costs when the record was released, why not price the CD sensibly?
I would like to answer several questions that were raised in this discussion.
(1) How does a CS person learn biology? I recommend "Recombinant DNA, A short Course", as an accessible (Scientific American style) introduction to the cloning breakthroughs and discoveries that lead to genome science.
(2) How does a CS person learn "Bioinformatcs"? I strongly recommend "Bioinformatics - Sequence and Genome Analysis" by David Mount as an accessible and extremely comprehensive survey of current approaches in Biological Sequence Analysis.
(3) Why do Biologists use Perl? Much of the information Biologists want is on the WWW, and Perl's LWP makes it extremely easy to get it. We don't use Perl for sophisticated text analysis (similarity searching, motif searching, etc) because the algorithms that are appropriate are typically not exact (or even regular expression) matches. But it's difficult to beat Perl for getting stuff off the WWW.
(4) Why do Biologists use Flat files? Several reasons - (a) the most useful information is sequence information, and it can be read much more quickly out of a flatfile (esp. one that is memory mapped) than a DB; (b) flat files solve some versioning problems that DB's make very complex and slow. (c) Most data providers only provide flatfiles. This will change, however, over the next 2 - 3 years, mySQL and postgresQL are moving into biology labs.
It is very exciting that Bioinformatics has high visibility now, and many people with CS background are considering bioinformatics problems. Unfortunately, many of the introductory books on bioinformatics (particularly the O'Reilly books) do not adequately present the substantial foundations of bioinformatics that have been build over the past 15 - 20 years, and some newcomers are mislead into believing there are simple problems looking for a few good programmers. Most of the simple problems have been solved; many of the complicated problems are challenging not because we do not know enough CS, but because we do not know enough biology.
The Silicon Valley article is a bit misleading, and doesn't accurately reflect the range of distribution alternatives being used for Bioinformatics software. It is certainly true that many Universities claim ownership of computer software copyrights, but it is important to appreciate that there many levels at which the implementation of these policies is decided. For example, both the WU-BLAST and the HMMer packages were developed by researchers at Washington U. in St. Louis. WU-BLASTbinaries are available to academics after an appropriate license is signed, and licensed commercially. HMMer i is available under the GPL but a commercial license is also available.
Likewise, the FASTA package, can be freely downloaded by both academic and commerical users, but must be licensed from the U. of Virginia to be redistributed. This has allowed the software to be widely used by researchers and also incorporated into commerical packages.
As a Bioinformatics researcher and software author, my goal is to have my research and software be used as widely as possible. This improves my ability to obtain future external funding, to get my papers cited, etc. etc. Even at universities like Wisconsin and Stanford, which derive enormous sums from IP licensing, these funds are less than 10% the value of NIH and other external funding. Thus, it is not hard to argue that software licensing policies should maximize the likelihood of external funding, and the widest possible distribution (though not necessarily GPL) is likely to have the greatest impact and long term benefit. (Moreover, once software becomes widely used, it is much more valuable commercially.)
Thus, while a university's Vice-President for Research may be interest in IP licensing, a Dept. Chairman may be more interested in faculty success in obtaining external funding, and a broader software distribution.
Before the Bayh-Dole act, the federal government owned the intellectual property produced by grants and contracts, and it was exceptionally bad at providing the patent and other protections necessary to get discoveries into the market place. Drug companies do not develop and market drugs if they cannot control the patent. Likewise, some software companies may not make the marketing and support investments necessary to commercialize a product if the product is available as open source.
Much of this discussion assumes that the "discovery", or IP, has substantial inherent value that can be exploited without additional investments in support, marketing, etc. For specialized fields like bioinformatics, this is simply not true - the program may provide important new capabilities, but it will have little commercial value without appropriate packaging, integration with other biological information, training, and support. The universities do not do these "commercial" tasks, neither does the government. Companies do this sort of thing, but not without some guarantee (e.g. an exclusive license) that limits their risk.
This comment reflects a widely held misconception about how research at public and private universities is funded. Much less than 10% of the research in the physical or natural sciences, or computer science or engineering at major universities (public or private) is funded by the institution. Almost all research is funded by federal agencies (NIH, NSF, DOD), and internal funds for research often come from indirect cost recovery from the federal grants. States and private endowments simply do not pay for research. (At fortunate institutions, they pay for buildings.)