This was more a note that, when the paper is accepted, those identifying things go back in. The proposal to have papers be anonymous would have to leave them back out, stripping the paper of important context, and making open science a bigger pain than it already is. Which is kinda contrary to the goal.
In practice, someone who really wants to break the double-blind veil probably can. But I don't think that's a problem, as very few people are consciously biased, and will actively seek out such identifying information. By removing names and affiliations, and pretending all of our work was done by someone else, I think that's enough to help people avoid unconscious bias.
Makes me wonder if you could implement the exporting of a "review" version in LaTeX, similar to a debug mode -- set a flag, and any content with an {ID} wrapper around it gets left out or is substituted for something anonymous.
While you can say "Remove obvious links", to many experts $Topic in $Location is enough to allow a pretty good guess of whose lab its out of. Or the name of a long running study. And I'm not going to maintain a whole separate Github account to anonymize code distribution.
"(AUTHOR et al.) performed a small scale lab study. (ANONYMOUS SUBMISSION) performed that exact same study, on hundreds of people, and expand on the work of authors..."
- GitHub, FigShare, institutional repositories etc. for code and data
- Information on the location of the study
- Statistical packages used (hope you didn't author the one you're using)
Of course, anonymity will be shattered the moment anyone talks about their work at a conference. Furthermore, several systematic reviews and meta-analysis studies I've worked on benefitted greatly from being able to email the authors directly with questions.
I'd also be concerned that people will mentally assign things to a "Big Name" in the field anyway. "Ah, another paper on HAART. Must be out of so-and-so's lab..."