dcsommer

23 ms·

211.
by dcsommer 12y ago
We use a very different model actually. Since spawning OS threads is expensive, we opted for the popular nonblocking-IO approach. Each worker thread (usually 1 per core on the CPU) is given connections in a round robin fashion from the list
212.
by dcsommer 12y ago
Maybe not today. We haven't open sourced a configurable proxy yet, so you wouldn't be able to do your redirects and some of these other features without writing C++ code. If you're interested in how HTTP frameworks are design
213.
by dcsommer 12y ago
This is similar to the Apache License, which allows developers to use the project with the confidence that we grant a license to any patents that may affect the project. This is a grant we use for all of our projects and is not anything spe
214.
by dcsommer 12y ago
Well besides the fun of hacking around and building little HTTP servers, I could see this being useful if you want to save money by running fewer instances of your HTTP service. For instance, if you have a widely deployed python webservice
215.
by dcsommer 12y ago
I'm flattered, but I'm just an author. Proxygen is the work of about a dozen people over 4 years at Facebook.
216.
by dcsommer 12y ago
Websocket support isn't out yet unfortunately. It's something we hope to get to soon. Our reverse proxy uses proxygen and does TLS termination too, yes.
217.
by dcsommer 12y ago
Yup, we still use Proxygen (the library) in our reverse proxy. Maybe some day we'll be able to open source the reverse proxy too, but it's pretty deeply integrated with internal FB code right now so that's tricky. We already
218.
by dcsommer 12y ago
It's requests per second (averaged over a 60 second test run).
219.
by dcsommer 12y ago
Hey there, I work on Proxygen at Facebook. I'm happy to answer any questions you have about the project.
220.
Introducing Proxygen, Facebook's C++ HTTP Framework(code.facebook.com)
4 points by dcsommer 12y ago|0 comments
221.
by dcsommer 12y ago
Reasons being increased performance and security.
222.
by dcsommer 12y ago
Yes there is. Knowing when a client has a resource cached already is an important part of server push. There is indeed a risk of over pushing.
223.
by dcsommer 12y ago
The problem is that opening new connections is horribly inefficient. The 2-3 round trips (TCP + SSL) required to set up a new connection and the ensuing slow start phase significantly delay the request, and thus the response. It is much m
224.
by dcsommer 12y ago
The real world impact is incredible, not small. Opening new connections is horribly, horribly bad for internet performance. http://www.chromium.org/spdy/spdy-whitepaper
225.
by dcsommer 12y ago
What in particular is bad about HTTP/2? Complexity can arise in protocols for multiple reasons. In this case, security and correctness are more culpable than anything else. That is, if you want to design something similarly performant,
226.
by dcsommer 12y ago
This is a ridiculous claim. The spec had continued to change and it is entirely understandable that there will be bugs. This is what interop testing is for. I've worked with the Google folks on the httpbis group and they are committed
227.
by dcsommer 12y ago
Stock SSL on iOS doesn't have NPN, which is needed to negotiate SPDY on the internet.
228.
by dcsommer 12y ago
> Go and Scala have powerful type inference, making explicit type declarations very rare. In C++ and Java everything needs to be declared explicitly. I guess they forgot about C++11's auto?
229.
by dcsommer 12y ago
Its unlikely musicians used equal temperament 3,400 years ago. This is a guess, and that fact is reinforced by the other "interpretations" referenced in the article. It's fun to imagine, but this not a "Rosetta Stone for
230.
by dcsommer 12y ago
I'd like to plug the policy my friend's startup is using. Unlimited vacation plus mandatory vacation. https://sourcegraph.com/blog/mandatory-vacation I think this strikes a good balance of establishing a cu
231.
by dcsommer 12y ago
> Poker is a game of skill with an element of luck. It is not a game of chance. It depends on the level of the players. For beginners and casual players with less strategy, the game is much more about luck.
232.
by dcsommer 12y ago
Managed languages is one way to go, but they are not really appropriate for performance critical libraries like OpenSSL in my book. You can get safety without being managed in modern languages if the compiler does the safety proving for you
233.
by dcsommer 12y ago
The language hasn't reached 1.0 quite yet. Give it 2 years and then ask this question.
234.
by dcsommer 12y ago
Another great entry. Keep 'em coming!
235.
by dcsommer 12y ago
I like this article. It's short and easy for the target audience and has enough content to let C++ developers start to understand Rust. It also hints at and leaves unexplained other features in the language, piqueing the curiosity of t
236.
by dcsommer 13y ago
People have documented this many times. It boils down to surprising features of the language. http://johnkpaul.github.io/presentations/empirejs/javascript... https://speakerdeck.com/felixge/ja
237.
by dcsommer 13y ago
Seems like you could apply some image classification techniques to automatically identify uncharted cities given access to all the satellite data.
238.
by dcsommer 13y ago
This is not true.
239.
by dcsommer 13y ago
Panspermia only interests me in an intra-solar system context. I imagine the probability of ejected rocks from an asteroid impact in one solar system reaching a body in another solar system is significantly lower.
240.
by dcsommer 13y ago
I'm curious if that is a function of how Facebook works or if that is what naturally happens when you see everyone you know online doing fun things (which would apply to every social network of that size).

More ›