CCL:G: Science code manifesto



Hi! A couple of people have suggested that explaining and providing details of software code is like providing details of the design of a piece of equipment used to perform a measurement. I don't think this is accurate. Experimentalists have to justify and validate the experimental approach, not the design of the equipment. The same practice should be part of manuscripts using specific QC code for the first time. Dave

On Wed, Oct 19, 2011 at 3:00 PM, Laurence Cuffe cuffe^_^mac.com <owner-chemistry*ccl.net> wrote:


On Oct 19, 2011, at 03:41 PM, "Jaroslav Kalinovski kofeinu:+:http://hotmail.com" target="_blank">hotmail.com" <owner-chemistry- -ccl.net> wrote:


Sent to CCL by: Jaroslav Kalinovski [kofeinu-$-hotmail.com]

Hi,
I really do not understand... for me, the source code I am writing is like an equipment in the laboratory. Do you really need someone else equipment to make a proper review? Maybe in this case it is easy to send the code but rules are the same. Do you ask for particle accelerator when reviewing paper about experiment with one?

If someone is having troubles with reproducing my results, one always can write to me and simply ask for the code but I do not feel I HAVE TO publish code explicitly. In the end it is my property, I can describe algorithm, points of theory but why should I give the code? No one is watching at the hands of experimentalists while reviewing their papers.

I think people are forgetting that code is just a tool not a research result.

It is true that relying only on reputation is not the perfect way but it works for other disciplines. In the end we always have to put on some trust in authors.
That is all for me.

Best regards,
J.Kalinowski
 
And that's exactly it, you don't have to publish your source code, but you should be prepared to let someone look over it if they ask.
Also you don't have to let them use it, you can protect that by copyright, but people should be able to look over it and satisfy themselves that it should do what you say it will.  The right to look should be free, the right to run or license the code can be whatever you think the market can bear.

The call for openness is in the context of devising code to analyze large data sets in the context of climate change. In this context we should not be comfortable with a researcher telling us that they wrote a program to extract the key data, who then asserts that they wont let us see how it works.

 I don't expect a reviewer to check the code implementation of every computational development that apears, this would be a very onerous task, and would not be realistic. However if a problem arises where a number of programs claim to be calculating the same thing, but get different results, then I think the scientific community would be best served if they knew that the source code was available for their perusal.

There may be a case for certain restrictions to apply, but these restrictions and limitations should be clearly flagged in any resulting scientific papers.  A very hypothetical example of this might be where a researcher was working for a pharmaceutical company which had developed proprietary software for developing leads in drug discovery. If the researcher wished to pubish details of research in which a number of potential drug candidates had been developed by applying the software to orphan drug discovery, without giving details of precisely how they had been developed then it might be appropriate to flag this and publish the leads as a matter of public interest.

One final point, open access to your code does allow ones rivals to see what and how you have written it, but it also allows you to examine their code and identify cases where they have stolen your IP.

All the best

Laurence Cuffe

---
Laboratory of Physical Chemistry
University of Helsinki
---
Logic is a systematic method of coming to the wrong conclusion with confidence.


----------------------------------------
> From: owner-chemistry- -ccl.net
> To: kofeinu- -hotmail.com
> Subject: CCL:G: Science code manifesto
> Date: Wed, 19 Oct 2011 07:50:08 -0700
>
>
> Sent to CCL by: George Fitzgerald [George.Fitzgerald[A]accelrys.com]
> As a member of a company that makes money from selling software, I probably have a different outlook on this than most CCLers. But I have one very practical question: as a reviewer, do you really have the time and expertise to review 1000s of lines of source code? I find that properly reviewing a paper already takes several hours. From experience I know that reviewing somebody's source code can take days.
>
> Can anybody give me an example of what you'd even look for in the source code? I'm thinking back to, for example, Peter Gill's 'PRISM' method for Gaussian integration, or Benny Johnson and DFT analytic 2nd derivatives. Are you claiming that those papers shouldn’t have been published without the reviewer reviewing the code?
>
> -george>
>



-= This is automatically added to each message by the mailing script =-
the strange characters on the top line to the - - sign. You can also


E-mail to subscribers: CHEMISTRY- -ccl.net or use:
http://www.ccl.net/cgi-bin/ccl/send_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST- -ccl.net or use
http://www.ccl.net/cgi-bin/ccl/send_ccl_message