Hacker Newsnew | past | comments | ask | show | jobs | submit | pacalleri's commentslogin

Similar experience last week. But tbh I'm using the free plan so I wasn't expecting too much from them. What it worked was to use nginx rate limiter aggressively, parse logs and deny top ips with nginx. Because all traffic comes through CF I wasn't able to use iptables for blocking


If you can thwart it with your own nginx, then it can’t be much of an attack. Cloudflare is one of your only hopes against a volumetric attack especially when paying $0.


Cloudflare’s has a free rate limit feature, btw. Not as configurable as nginx but it’s nice to not have the requests touch your server at all.


Same bad experience here with socket.io.

I switched to Faye (http://faye.jcoglan.com/) an implementation of Bayeux protocol programmed by James Coglan. I'm using the node version of the library in my games for 4 years with excellents results.


How are you scaling it? I'm using the Ruby version along with eventmachine but max concurrent connections was 1024.


sounds like you're hitting the ulimit for your server process, rather than a issue with Faye itself.


Can someone clarify why is nobody publishing the complete list of names or the raw data files?


While I was reading came to my mind the Borges's short story "Funes the memorious". It's about someone who can't forget any detail. He remembers absolutely all the things and the infinite instances of them through the time. At some point of the story Borges conjectures: "I suspect, nevertheless, that he was not very capable of thought. To think is to forget the difference, to generalize, to abstract. In the overly replete world of Funes there were nothing but details, almost contiguous details."


Great passage. As a digital artist who works with fractals, that really resonates with me. Visual fractal detail quite often converges to visual noise (and looks remarkably like a noise function as expressed on e.g. a TV set). I usually need to remove or de-emphasize that noise in order to clarify the direction and abstract intent of the work.

One of my favorite films that works along these lines is the 1998 Japanese film "After Life," in which a small party of workers attempt to recreate others' memories with very basic film studio equipment. I absolutely treasure the loss of detail in the various recreation scenes, and the way it suggests that there is actually a satisficing point at which we might realize, "yes, I'm actually reliving that memory right now." So I agree with Mr. Graham's conclusion that technology can bring this about.

On an unrelated note, PG's essays always bring to mind the Meyers-Briggs INTJ type. Essays about the annoyance of accumulating "stuff", a focus on abstract / intuitive learning styles, and clever writing which quickly establishes a theoretical framework which is then thrown against the world's (audience's) experience, rather than starting from first principles hoping to eventually reveal a framework as others might do. His seems to me very much a "systems thinker" approach.


I've noticed that whenever truly original thinkers encounter a problem, they'll quickly establish a workable model—even if it's known to be flawed or wrong—just so they can begin testing it “against the world's experience.”

(I had no idea this style of thinking was associated with INTJ types.)


My AI prof once wrote on the board "Learning is generalization". It makes sense: if the only way you're reasoning is by attempting to retrieve things from your cache, it's rote learning and you're not able to deal with situations you haven't encountered before. And even if your cache is super-extensive, all you're doing is overfitting.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: