In my last post, I briefly covered the ‘share or not to share’ debate involving non-commercial software. In this post I’ll delve deeper into the issue by discussing how commercially available research software further complicates the situation. I’ll focus on perhaps one of the most controversial conflicts in the chemistry software: Gaussian Inc. vs Banned by Gaussian.
In the 1950s and 60s Prof. John Pople (1998 Nobel Prize winner) and his research group at Carnegie-Mellon University were focused on the development of ab initio quantum calculation methods. The group incorporated Gaussian orbitals – rather than Slater-type orbitals, which were more computationally intensive – into a computational chemistry program for molecular electronic structure calculations. The program, Gaussian 70, was released as open source software through the Quantum Chemistry Program Exchange (QCPE) in 1970.
In 1987, Carnegie Mellon University was issued a software license for the program and, ever since, it’s been developed and sold by Gaussian, Inc. Prices (pdf) for the Gaussian software package range from $2,500 for a single computer to $35,000 for an institution-wide license.
Gaussian was initially used only by theoreticians. However, as I mentioned in my last post, the continuously increasing power of personal computers as well as the addition of a user-friendly interfaces have made the software so accessible that even a computationally inept synthetic chemist (like myself) can perform high level ab initio calculations with a half dozen mouse clicks.
Gaussian is an important tool for many chemists, but it’s has also been a center of controversy. Since its commercial release a number of individuals and institutions have been Banned by Gaussian (BBG), which means they are prohibited by Gaussian Inc. from purchasing or using any version of Gaussian software.
In 2004, a website was established – bannedbygaussian.org – to bring attention to Gaussian’s policy of banning competitive researchers. The website has a brief history of Guassian software, what it means to be BBG, what it takes to be banned, and a list of researchers/institutions currently banned (Interesting side note: Prof. Pople cut all ties with Gaussian in 1991 and was subsequently banned for modifying the software).
According to the BBG website, bannable offenses include: 1) Publishing comparisons of Gaussian’s performance (computation time, result, etc.) to another software package, 2) announcing that new software is faster/better than Gaussian, 3) modifying Gaussian software, and 4) sharing Gaussian with those that are BBG.
The website’s creator is anonymous so it is difficult to verify the validity of the claims/content of the site. However, if you are willing to accept my submission to the Journal of Anecdotal Evidence, I feel confident that the animosity expressed in bannedbygaussian.org is real based on conversations I’ve had with several people within the theoretical community.
The controversy over Guassian’s policy of banning competitive researchers has even made its way into the pages of Nature and CandE News. Also signaling increased inattention to the issue, the World Association of Theoretically Oriented Chemists (WATOC) passed a resolution in 1999 opposing the restrictive license practices of companies like Gaussian Inc.
In the previous blog post I listed several concerns about the practice of publishing research results without sharing the software and source code used to calculate the results. These concerns not only hold true for commercial packages like Gaussian, the act of banning competitive users also raises issues like:
- Hindrances in peer review. As stated by the BBG website, “Many journals, including JACS, will not publish computational papers unless their results can be verified.“ If researchers are not allowed to double check Gaussian-calculated outcomes, they cannot properly review research articles that use this software. Prof. Mark Gordon (Iowa State University) who is BBG says that he can “only review about half of the 40 to 50 papers [he] normally reviews in a year.”
- What do we value? Profits or knowledge? It may be a harsh reality, but it’s unlikely that anyone will ever make tons of money selling computational software. The question then becomes: what is the value of protecting Gaussian’s software to such extremity? The measures used to protect the software are off putting to many and insulting to the scientific community. The BBG website describes the scenario quite simply as a situation where “commerce trumped science.”
- An alarming expectation for user loyalty. Peter Gill (a former postdoc of John Pople, current professor at Australian National University, and resident of the BBG list) compares the banning policy to “ a worker from Ford not being able to buy a Chrysler car.” In other words, by saying that their users cannot make public comparisons between their products and competing products, Gaussian is demanding a level of customer loyalty that would be ridiculed and rejected in any other market. Could you imagine if critics were not allowed to compare movies from different production companies? Or comparisons of macs and PCs? Of Coke and Pepsi?
What does Gaussian Inc. have to say about these issues? They have a “Comments on the ‘Banned by Gaussian’ Website” page where Michael Frisch, a former graduate student of Prof. Pople and current President of Gaussian, Inc., addresses some of the “half-truths and outright misstatements.” The URL – http://www.gaussian.com/g_misc/silly.htm – implies that the company does not take the controversy very seriously. Yet, they do respond to the criticisms. Here is an overview:
- There is nothing to trust in anonymity. Because the website is anonymous, Dr. Frisch suggests that the creators do not have the “courage to stand by their malicious statements, nor the integrity to admit their own interests in competing programs….”
- No universities are banned. There are a number of Universities listed as banned on the BBG website. However, Dr. Frisch insists that Gaussian has never refused a license to an entire university. However, they do acknowledge two occasions where they’ve refused a license to an organization (unnamed) on the grounds of flagrant user agreement violations. They also openly acknowledge the banning of research professors and their group members.
- Other software packages can double-check Gaussian’s work. Dr. Frisch pointed out that, for most calculations (emphasis added by me), other software packages can use similar functionals and basis sets to obtain the same results as Gaussian. As a result, the actual ACS standards are fulfilled (emphasis by Dr. Frisch). Additionally, all methods and algorithms are published in peer reviewed journals.
- An unfair standard of sharing: Dr. Frisch explains that most researchers do not immediately share their source code upon publication. Why does the BBG website specifically attack Gaussian Inc. for not sharing their source code?
- What it takes to get banned. They describe the banning criteria listed above as “downright silly.” Unfortunately, I have to admit I don’t entirely follow Dr. Frisch’s reasoning. He says that bannedbygaussian.com misrepresents Guassian when saying that “applying our software to chemical problems and using or publishing the results would get anyone banned.” I’m at a loss because I don’t see where that claim is made on the BBG website. Instead, I believe they are saying that publishing comparisons would result in a ban.
Much to my disappointment, nowhere in Gaussians rebuttal do they 1) describe what actions will lead to an individual being banned and 2) provide a justification for the bans.
I am not a theoretician and do not fully understand the intricacies of computational software. However, my strong desire for openness and collaboration in science leaves me critical of Gaussian, Inc. for the practice of banning.
With that said, I would like to bring everyone’s attention to science friendly, open source alternatives to Gaussians computational software like GAMESS, PSI, NWChem, MPQC and many others. My one critique about many of these open source projects is that their user interfaces are usually text based or you need a secondary program to use as a GUI. Because of these difficulties they are not nearly as user friendly as plug and play commercial packages like Gaussian/GausView, Spartan, and others. That does not mean that they are hard to use, it just means that there is a notable activation barrier to learn how to use them, an investment that some non-theoretical chemists are unwilling to make.
In closing, I need to share one last lingering thought: Will anyone from Gaussian read this post? Is there a chance that I’ll be banned for writing it?