What I found more interesting than Pachter's analysis of the specific failings of the Barzel-Barabási paper is the more general implications for our scientific process. I have two parallel trains of thought:
1) We all know that we have scientific "celebrities" and that their work is often over-represented in high-quality journals (while dissenting viewpoints are often suppressed). There is nothing wrong with critizing the work of the celebrity du jour (or in this case, maybe, du décennie), but framing is everything and you will find more success in your scientific career with constructive and respectful criticism without resorting to ad hominem attacks such as "the emperor has no clothes". No one benefits from a bloody cage match.
2) Network science (like many interdisciplinary fields) has problems because many of its practitioners lack extensive knowledge of the systems, experimental challenges and core research methodologies of the fields or contexts to which they apply their developments. Success in such interdisciplinary fields is often incentivized and measured in a very narrow way and not associated with the primary goal of science -- useful and meaningful advances in our collective body of knowledge. You want to be at the top of your field, but presumably (or hopefully) you have more noble aspirations -- you want your work to have real meaning. But this lofty goal is challenged by the rapidly growing body of scientific work and the fracturing of science into siloed sub-fields. For example, finding the right set of peer reviewers for an interdisciplinary submission becomes increasingly challenging. Consequently, mistakes will be made, missed in the peer review process, and published. Dissenting opinions and followup work need to be given equal representation.
I'd amend this to "tell me about an academic publication that you've been a major contributing author". Depending on the field and circumstances first author doesn't mean the same thing.
The basic technology for this has been around for years (e.g., the logitech digipen), though I think this latest iteration is the first to add pressure sensitivity. I've always wanted one of these (back when I was getting my physics PhD, I would have killed for this), but everything out there to date seems to be the same rehashed product that is doomed to be abandoned shortly after its release. Here's to hoping wacom does it right.
I suppose it's easy to assert that a site is immune to ES when it's in its infancy-- the argument made here is that HN's niche hardcore basis protects it, but the same argument could have been made for reddit 5 years ago, a case which they curiously use as a token example of ES. I think the takeaway here is that only intentional and deliberate measures to protect against homogenization of a community will work -- and these are often less egalitarian by design.
Sorry, so the argument is: other research costs are expensive in some fields, thereby high spending is prestigious, thereby access to publications ought to be expensive so that fields that lack more direct expenses can seem comparable ? This doesn't seem to make sense, because fields where direct research expenses are high also require access to publications. Maybe I'm missing your point? If so, please clarify.
The advantages and disadvantages the author cites seem more pertinent to his own idiosyncratic preferences than to more general features one might look for when doing interactive data analysis. They also seem easy to address. For example, an hour of time spent building a few quick functions would address most of his complaints about Python. I've personally used python, matlab, R and Stata in my research and view the first three as about equally capable. In my opinion Stata is less comparable to the others as it is more a wysiwyg collection of tools and functions. Matlab has good support for large data sets via memory mapping, has mex extensibility for building your own fast functions and is very good for interactive plotting, but doesn't produce publication-quality finals. Python is great for no-niggling fast idea to functioning execution and can push data into matlab is mlabraw. R has well developed stats packages and a huge user base. I disagree with the author regarding documentation for R--maybe he is right for the core, but depending on the package you may have trouble finding documentation beyond a man page. Ggplot is excellent but eccentric.
Topological properties of an individual (such as network degree or betweeness centrality) are not necessarily predictive of influence. The author acknowledge this and suggests observation of peer response to messages sent by FB on behalf of the user. What is really necessary here to determine influence is a randomized trial. See for example the methods described here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1564856
(full disclosure - I am a co-author of this work and of some upcoming work that describes techniques of identifying influential and susceptible users in social networks).
Beyond managing your identity/s, the most compelling notion this brings to my mind is that sites should reduce transmission of redundant data: You specify the structure once then only the data is transmitted (what someone else referred to as "like iOS apps"). To an extent we are already moving in this direction with asynchronous update technologies, but rather than a next-gen browser, it seems like we need a next-gen language that places a primary emphasis on reducing wasteful transmission of redundant information. E.g., On iOS I update my FB app when there is a change to its structure and then I do so exactly once; Why am I re-downloading its structure in my browser window multiple times a day?