Copyright Status of Thermodynamic Properties? 154
orzetto writes "I work at a research institute, and programming models of physical systems is what I do most of the time. One significant problem when modeling physical processes is finding thermodynamic data. There are some commercial solutions, but these can be quite expensive, and to the best of my knowledge there are no open source efforts in this direction. In my previous job, my company used NIST's Supertrapp, which is not really that expensive, but is written in Fortran, and an old-fashioned dialect at that. As a result, it is a bit difficult to integrate into other projects (praised be f2c), and the programming interface is simply horrible; worse, there are some Fortran-induced limitations such as a maximum of 20 species in a mixture. I was wondering whether it would be legal to buy a copy of such a database (they usually sell with source code, no one can read Fortran anyway); take the data, possibly reformatting it as XML; implement a new programming interface from scratch; and publish the package as free software. Thermodynamic data is not an intellectual creation but a mere measurement, which was most likely done not by the programmers but by scientists funded with our tax money. What are your experiences and opinions on the matter? For the record, I am based in Germany, so the EU database directive applies."
It will be a very difficult project (Score:5, Interesting)
I can't find my copy of Supertrapp at the moment, but as I recall there is some strange wording in the license. It's definitely NOT public domain as asserted by the uninformed.
It's also not tabulated data. It's a collection of equations and empirical constants embedded in what may be the worst code I've ever seen.
It may be easier to track down the original papers and work from those, though that too is difficult as lots of the original work was published in obscure journals.
FWIW I am very comfortable w/ FORTRAN and prefer it for serious numerical work (default choice is C). I'm also quite skilled at interfacing FORTRAN to other languages.
I'm interested in working on such a project and have quite a bit of experience w/ the problem, though only limited experience w/ Supertrapp because it is so bad I tended to avoid using it unless I absolutely had to. Please send me an email so we can discuss more. rhb acm.org
Reg Beardsley
Re:It Probably Wouldn't Be Legal (Score:5, Interesting)
I do not know about this exact database, but many scientific databases are hand-curated and extensively reviewed. Many do not include every measurement published in the literature, but carefully and judiciously select those data points deemed, by expert opinion, most reliable. Thermodynamic databases do not contains "facts" per se, but measured data points which may or may not be close to the facts. The editing and review process, which is quite an investment, does often create a solid foundation for copyright. These databases are not just a routine business, like a reformatted dump of the data from a telephone company.
Re:NIST - Public Domain (Score:2, Interesting)
It appears that there is an exemption to the public domain status which applies here:
15 U.S.C. Â 290e authorizes U.S. Secretary of Commerce to secure copyright for works produced by the Department of Commerce under the Standard Reference Data Act.[8]
Re:FORTRAN (Score:3, Interesting)
Definitely copyrightable (Score:3, Interesting)
Empirical models of thermodynamic properties are definitely protected by copyright. There is a high-value market for these models, and different models of the same thermodynamic process will evaluate differently so it is a valuable creative product rather than a mere description of reality. For fields where tiny improvements in efficiency generate big cost savings, you want to use the most accurate model available where "most accurate" will be a function of the use case.
Thermodynamic property models are not measurements of reality, they are mathematical models of a physical process derived from empirical data. They are what you use to predict reality when it is not possible or practical to measure it. Turning the empirical data points into continuous functions is a creative step and the value of the creative step is in minimizing the divergence between the model and reality over as broad a range as possible. There are companies that specialize in producing and selling ultra-accurate thermodynamic property models.
Re:Thermodynamics (Score:3, Interesting)
I think that the biggest problem isn't intellectual property, but the people who administer it. I don't think that the demand is particularly great. As such, there isn't a great incentive to release it freely. There are costs to administering such a large DB. Furthermore, nobody wants their name on a database of all the fundamental properties because in that data there are bound to be mistakes.
You are looking at the liability issue for the creator/admin, the supply side. The bigger liability problem is on the engineer, demand side.
Something that is missing from this discussion, is some Chemical Engineer specific knowledge that I can attempt to provide. The whole point of a "steam table" and similar products like discussed here, is there is no accurate formula for vapor pressure at various temps. The simplistic linear equations taught in high school don't work at the extremes, or don't give accurate enough results to design a safe and profitable plant. So, more than a century ago, physicist / chemist / engineers started making lab measurements, and selling graphs and tables of data. The modern version of that product is the expensive computer models discussed in the article, which optimistically try to answer any input conditions with correct and continuous answers based on a mixture of theory, optimism, and some distinct individual laboratory measurements.
Because the data model is used to design multi-million dollar plants, and because the only way to verify the results is very expensive lab work, and is therefore often glossed over, a mistake in the data model could be a multi-million dollar mistake, assuming the losses are purely economic and no human victims. The creator/admin probably was intelligent enough to release under a license that removes all liability for data errors. The end user engineer will not be so lucky.
On one far extreme of the provability / testability spectrum, you've got yet another word processor, where if the screen doesn't match what you typed in, literally a trained gorilla could figure out the word processor is broken, and act accordingly (throw poo at programmer? The more things change, the more they stay the same) Or maybe a crypto hash where a hundred programmers can write it in a hundred languages and all the outputs better match for a given input.
At the other extreme of provability / testability, you've got a Chem-E basically having to take the program output on faith that it's correct. The program says the pressure of supercritical steam at 700 K is 230 atm, I know that is somewhat above critical temp and above critical pressure, so the best I can do is "sounds about right to me"? So specify plant components based on a 230 atm environment (adding appropriate safety factors, etc) Now steam pressure is old stuff, boring, and everyone knows about what to expect, but using really weird stuff under really weird conditions, who knows what crazy output from the data model might slip past, resulting in a disaster?
Despite the dangers, it would be great for education, and cheap experimental research/simulation, even if it would be too legally dangerous to use in formal design work.
Re:It will be a very difficult project (Score:3, Interesting)
Items created/published by the U.S. government are in the public domain at least in the U.S. I'm not sure if the rights are granted abroad as well.
However, items created under a contract to the U.S. government may or may not be in the public domain. There's a section of US law that companies have to invoke in their contract or the software license regarding U.S. government "limited rights" to keep their code or other work private. On the other hand, NIST DATA shouldn't be copyrightable in any case, although companies still like to test the theory that their databases are copyrighted fairly frequently.
I am not a lawyer. Even if I were a lawyer, I'm not YOUR lawyer. The above discussion is mostly intended to point out that you're more likely to get a better deal for data out of the government than out of a private company. If you need pre-built software from a private firm, you're likely to be tied into a license agreement that affirms copyright over the data or its storage format (or both), which could be a source of disagreement with the software vendor if you decide to follow your plan of extracting the data for your own use.