I'm in the process of moving away from a static site for my website.
Simply put, a web interface is often more convenient so I can update things easier on the go, and it's good to have a place to host experiments too.
I have used various static site generators but none of them seemed to be significantly less work to get going than a small django app on heroku. Though I've been using Django for a few years, so there is simply no learning curve left.
It's surprisingly easy, though I believe the tests expect one to be configured (I haven't checked precisely why, or if it can be configured not to expect a database).
"Social networks -- real social networks, not the online instantiations of them -- work best when the groups are relatively small, Dunbar's number is respected, and there's a level of insularity around any given group."
Agreed, and to add to your point, group identity is also a strong motivator for close-knit networks.
Forums are a good example of this. It's partly why I think communities (and the software they use) centred around a specific interest or activity will, for certain kinds of community, ultimately win over what we now call social networks.
It's the tools that are lacking: at the moment it seems far easier to set up a facebook group than work out what a community needs and provide it without technical knowledge.
(I have an interest in this area because tools for communities are what I left my job to work on: http://microco.sm).
I agree with the main conclusion here, but it's a stretch to reduce technology choices to a simple "new or not" dichotomy.
Let's say you're writing a new web service in Java, because it has features aplenty and is also the language your team is most familiar with. You're confident the JVM is a platform you want to build on.
Now you need to:
1. Choose a set of libraries or a framework. Do you go for Spring or Java EE, or for something newer like Play or Dropwizard?
2. Choose a build tool. Maven? Ant? Gradle? Maybe we'll write some scala, so SBT?
3. Choose tools for deployment, config management, etc.
4. A database.
5. And so on.
All of these tools have different trade-offs. There are so many trade-offs that I don't think blog post comparisons (or whatever) cut it. And so you have the "magpies" who try and figure out some of these trade-offs for themselves by experimentation. (That is what, in my opinion, hack days and 20% time are for, not your new production system.)
But don't listen to me, we wrote our new web service in Go ;)
More seriously, it was a major decision and I couldn't possibly write a few hundred words on my blog to justify it. I may write a few thousand, though.
In that case, I don't know a number that can be used for that comparison. 18k is the total number of lines in the source code files; the SLOC will be different.
In hindsight, I should've left this number out. It was intended to give background to the overall program size, but it's a small detail that shouldn't be the focus of this discussion.
Just in case anyone has the same issue: on my device (Nexus 4) it took a while for the UI to become responsive when first opening the app. Patience solved this one :)
I've also heard good things about https://github.com/spf13/cobra and https://github.com/codegangsta/cli.