Articles are good for getting into a field you are not familiar with, especially when that field is obscure and lacks good textbooks.

The real problem is that we are continuing to support an industry that contributes nothing at all to the publication of scientific papers. We simply do not need publishers of any kind, regardless of open access, because we have a substantially better way to distribute articles: the Internet. Sure, we need archives of articles, but that is what university libraries are for. Sure, bandwidth is needed for distribution, but that is where P2P systems like BitTorrent come in (imagine a monthly torrent of peer reviewed articles for each field). The rest is already being organized without the assistance of publishers.

I was just thinking on the same lines. I work in the humanities and even if the article form remains, there is the possibility to radically rethink the journal model. The journal model is too closely tied to assumptions about print and distribution channels that no longer hold much value.

The function of research journals is to disseminate recent findings. In biology and medicine, it takes a while for a piece of published research to become accepted, as such recent findings are usually only of interest to other researchers. When something becomes accepted it enters text books and the like. Between the original papers and text books lie scholarly reviews and online databases.

Some examples: For individual genes, the established canon can usually be found online at places like Online Mendelian Inheritance in Man (OMIM) [0]. Furthermore, a lot of genetic data is now made available through what might be called "aggregators", such as Ensembl [1], which provides both a free web interface to explore genetic variation and also allows downloading of primary data. Similar resources exist for other types of data.

So, to answer your question, there are indeed a number of bioinformaticians who are trying to do exactly what you propose, some sources have become "aggregators of aggregators", for example Bioinformatic Harvester [3].

[0] www.ncbi.nlm.nih.gov/omim

[1] www.ensembl.org/index.html

[3] en.wikipedia.org/wiki/Bioinformatic_Harvester

Science is still a very creative (i.e. non systematic) process. To create new "scientific findings" you need new insights. These insights typically come from publications/conferences/discussions with other researchers rather than big machine readable datasets.

Most research articles I read can be boiled down to a paragraph of substance, with the rest of the space filled with introductions and things we already know. Agree about conferences etc. but I don't see how articles create discussion, since most journals do not even have comment sections, and even those that do, rarely have any comments. Articles are an arcane notion used to gauge researcher's output, until something new and more efficient comes along.

Ps. http://www.cell.com/neuron/retrieve/pii/S089662731300648X

Depends a bit on the area. You can machine-code results if the field is extremely static in its conceptual and epistemic framework, so it's just a matter of slotting new things in place. Some "industrial-style" science does work like that. But large areas don't.

Most of the articles I read have the majority of the actual research going on in the text, not in whatever the "results" might nominally be. For example, a programming language paper might prove a few theorems about type-soundness, but it's rare that a new type-soundness theorem is really want you want out of the paper; that's just supporting apparatus. What's more interesting is the design discussion, motivation for why this feature was introduced and how it relates to previous features, explanation of how it was implemented, discussion of variants and future work, etc.

Obviously this depends on the quality of an article, but there is more to the "extra bits" (i.e. that which is not plain data) than mere fluff. If you are very familiar with a topic (i.e have a PhD in precisely that area), the extra bits are not of much interest, though the methodology might be.

If you have some passing familiarity, then the introduction and discussion can help you a lot in understanding the implications proposed for the data. Most biomedical experimentation takes place in the context of considerable amount of uncertainty, e.g. relevance of the model used for human disease, etc, and more, as described in the article you linked to, so it does help many readers to have some kind of context.

The lack of discussion in biomedical articles is likely due to a fear of reprisals.

The "Open Data" movement is in fact underway. But adaption will be even slower there. When you look at the mess of formats and proprietary tools trying to lock you in in science, it becomes clear why that is. And there's a tendency in science to come up with ridiculously complex incompatible data models. First because Of the academic virtue of understanding it and second because nobody wants to fit a study to a data model but rather the other way around.

this sort of open data the poster one above suggested will always be problematic because the temptation will be to do meta-studies where you cross-compare datasets from experiments that 'look alike' while stripping context. It plays to a 'science is objective' fetishism, when in reality context is always subjective and context matters, deeply.

Presenting data as separate units in a publication serves the purpose of requiring actual cogitation and consideration before using the data. In a sense, all data are proprietary, and attempts to de-proprietarization of it is a mistaken endeavor. This is not to say that we shouldn't try to reproduce results, but we should be careful to acknowledge that the context is always different.

Science needs a short format that is frozen in a certain point (not very far at the life of the text). If you want to change something, great, but those two requirements are very important.

People would also benefit from a large changing body that reflects out current knowledge. But that was impossible until very recently, so every initiative on that line is new. Anyway, that second body can not replace the short frozen format.

If those descriptions makes you think about version control software, well, there are already a lot of comments talking about it.