The request is harvesting data from Facebook. For users with a tremendous amount of data, batch requests must be made, so it can take a while sometimes.
I can't say authoritatively for Python, but Twisted is probably what you need here (I have used both Node and EventMachine to do things like this, and I'm pretty sure Twisted is similar). The event reactor model will let you defer the request while it's doing things like performing external IO, so your application is freed up to handle other requests in the meantime.
Thanks! After spending the summer building the product, we're really excited to publicly launch Firefly. Your comment definitely captures our mindset - how many times do you call into a customer support line today and it just sucks? We think support is ripe for disruption and being able to see your screen instantly makes the lives of customers and reps that much easier.
Here at GraphMuse (invite engine for Facebook apps), we hadn't heard of a "Growth Engineer" until we really looked into some of the people interested in our product.
Turns out that Growth Engineers are actually respected "higher-ups." You said it yourself too. That said, I think few people are familiar with this job title outside of SV.
Was it Facebook who started this trend?
NOTE: I'd increase the line spacing on your blog, it's a bit tight.
Yes it was Facebook who started the trend as early days of the Facebook platform trained many of the best growth hackers in the Valley.
'Dan Yue' at Playdom, 'Joe Greenstein' at Flixster, and 'David King' at (Lil) Green Patch were a few Facebook pioneers.
But, it was 'Sean Ellis' (Eventbrite, Dropbox, Xobni, Catchfree) who Originally coined the term.
That's true hardly a few people are familiar with this job title outside SV.
I myself being a startup addict from India came across this title one night and couldn't stop dreaming the very moment 'how to be a growth hacker*.
Facebook provides you the user's list of friends, and each friend's list of mutual friends with the user. From there, we construct the friend graph, and then run our clustering algorithm on it.
You're right about Facebook not providing friends of friends. Supposedly they used to back in the day, but not anymore.
Makes sense, thanks! Didn't realize they provided mutual friends.
Great service, did a great job on my list of friends. I would love to use this, but having to send over user access tokens is a little scary, even over HTTPS. Have you guys considered licensing this? Would be cool as a heroku plugin...
We thought about licensing it, and we'd probably feel comfortable doing so. However, the infrastructure required to run this is pretty intense and customized. We're actually pushing a patent for the infrastructure/technology behind it.
If you'd like to brainstorm, feel free to shoot me an email at
tony@graphmuse.com
I have some ideas that are worth developing. In fact, we were using GraphMuse originally to determine entire fraternities and sororities. We listed 3,400 Greeks at UPenn in 3 weeks.