The current practice of printing archival scientific journals (as opposed to newsy journals such as Nature or Science) needs to be rethought.

What is the purpose of an archival scientific journal?

The fundamental purpose of an archival scientific journal (ASJ) is to referee the correctness of the presented scientific work. Many articles are very difficult to understand, except by experts in the specific field of the article, and other readers need the assurance that they are not being misled. Refereeing also helps prevent scientific fraud, but (especially in life sciences) often requires replicating the work—an expensive and time-consuming process.

The fact that the article is printed in an expensive, tree-destroying, difficult to access journal does not add to the dissemination of knowledge—arXiv does this successfully for free. Furthermore, it is not easy to do a full text search of printed ASJs unless it has somehow been digitized.

What is wrong with printed ASJs?

As I approach 80 years of age, my house shelves are full. I have no room for printed books, journals, CDs or LPs. Even the walls are full. Furthermore, since I have had my cataracts removed and replaced with trifocal lenses, I have difficulty reading printed documents. It is much easier for me to read on my Kindle, iPad, or desktop computer.

Understanding scientific articles is often difficult, even for an expert in the field. Printed ASJs often have page restrictions (e.g., Physical Review Letters) that force authors to leave out must intermediate steps in their equation presentations. An all electronic version of the work would have no such limitation, and make the articles much easier to follow, especially for students.

So, you may ask, why not just publish everything as PDFs? That solves my reading problem, and it does allow full text indexing, but it fails to take advantage of the newest (and future) technologies.

Suppose, for example, you want to replicate, check, or expand published data. This is impossible unless you contact the author and somehow arrange for a data and data format transfer. Furthermore, the metadata accompanying the data is critical to its use. I worked in the fusion energy field, and the world fusion experiments are huge, very expensive, and often international in scope. Knowing, say, the electron temperature is useless without also knowing all the other plasma parameters at the measurement time.

A very important and vital issue is that data are not necessarily fixed in stone; they can be dynamic. They depend upon instrument calibration, and mistakes can occur that cause existing data to change. Who has used these data? How do these users know that the data they used has changed? One example of this occurred (a long time ago) in the Magnetic Fusion Energy computers at Lawrence Livermore National Laboratory (LLNL). There was a mistake in the sine function in the third decimal place; to my knowledge, no scientific papers were changed as a result of this error.

What would a better idea for publication look like?

I believe that an ideal scientific publication would look more like an electronic Lab Notebook. It might have the following elements in it:

1. A comprehensive, understandable write-up of of the work with all of the intermediate calculation steps.
2. Computer code links to allow others to replicate the calculations.
3. Links to, or replication of the data used, together with its metadata.
   1. If I click on a data point in a plot, I would like a pop-up detailing the datum and its provenance.
   2. The ability to add my own data to a plot.
   3. The details of any fit to the data points.
4. Color used in all plots and tables to make comprehension easier.
5. The ability to rotate and zoom three-dimensional plots.
6. Enumeration of dead ends that were tried. Negative results can be useful.

To do this fully would require the solution of some major problems.

Foremost there needs to be a means to uniquely register data points—a tag that would be automatically applied to to new data sets that would allow things like data retraction and change, and also point to more details. Metadata are just as important as the data, and there needs to be a standard way of doing this that hopefully could work across fields. XML can organize and display such metadata, but its vocabulary needs to be developed. Scientists require that their data be used responsibly.

The fusion community has been trying to share computer codes for decades. I spent a year benchmarking a code with a Japanese colleague. Alas, such efforts are not supported financially or given credit as valuable research. Our codes agreed if we solved exactly the same problem. But this is non-trivial because most scientific codes rely upon assuming things such as the ranges of input variables, so that the numerical methods that are used are appropriate. Scientists are reluctant to allow others to run their codes without knowing that they will be used within the assumed parameter bounds. Face-to-face training sessions are often required to build up trust.

Moving to purely electronic publication of ASJs would save a lot of money, allow more rapid publication, and enable more widespread dissemination of knowledge. Not all of the above wishes can be carried out currently, but items 1, 3.3, 4, and 6 could be done immediately.