> > > Read about SEDA, I don't know if you have experience developing multi > > I am hearing about SEDA for the first time and it says it is a 'Staged > Event Driven Approach' and gives the details of how staging can help > in the Event Driven Approach. Now, we have quickly moved to a 'staged' > event driven from event driven. Pretty new concept, at least in the > Python world. How is twisted coming into picture here? >
I should admit that I'm hearing SEDA for the first time. I've had my fair bit of experience with Twisted. And in my experience I've never found a document which portrays Twisted as an answer for I/O concurrency problems. Twisted is not to parallelize I/Os but it has come to me as a framework that lets you write event driven code. No matter what you do, you will write blocking code in Python, but that's where Twisted comes for rescue. It helps you abstract how a particular blocking function is executed - maybe its threads, maybe its processes. Dont get me wrong, I'm not trying to say Twisted doesnt solve I/O problems, maybe it does. But in my expeditions (I/O was never a problem for us) I never used Twsited that way. It was more for Network communication and serving requests. Its funny because when I heard Call for proposals, I thought to myself - yeah, let me do a Twisted talk. I'm glad there are people in the group who are not just thinking, but executing that. Good Luck! /jeff On Mon, Jun 28, 2010 at 3:27 PM, Senthil Kumaran <orsent...@gmail.com>wrote: > On Mon, Jun 28, 2010 at 10:25:57AM +0530, Shiv Shankar wrote: > > Please read my presentation notes, I am not talking about Twisted, I am > > talking more about the 2 approaches, the second approach being used in > > designing twisted giving you a clear reason to use twisted. > > Okay, I read it again to get a better context of your presentation. > My understanding is, you are building up the requirements for a need > for asynchronous IO and then landing up with Twisted based approach. > BIND like software could be one example which could be chosen as > example out of any generic Server. > > > And, > > 0. If you just want to do Networking Software, asyncore is availiable in > > native python, why use twisted ? Where does IO comes in most networking > > software ? > > I did not properly get this question. In my understanding, you can use > asycore pretty most of the places where you can use twisted. Except > that you will have to do more. Write more code. Asyncore is different > from threading concepts for sure. But it is event driven. Twisted > gives you a framework with easy implementation of reactor pattern and > then gives you an ability to create differeds for handling blocking > requests in a non-blocking way. You could do the same using asyncore > with more work. > > > 1. Threads in any programing language is problematic, unless until used > > right.* Python is different, due to its abstractions.* > > - It's a pretty generic point. I am trying to understand what it is. > > > 2. For developing a concurrent system the most crucial thing is > > write to read ratio, Twisted is used for network programing because > > the ratio is low in most cases, and due to the async bindings for > > read/write it aids to concurrency. > > I got confused here too. Perhaps I am seeing it from a different angle > or I have not read the same literature that you have read. Are you > referring to IO operation by saying read/write. And also if you point > me to any literature on ratio of write to read (O to I? ) on > concurrency that would be helpful. > > > Read about SEDA, I don't know if you have experience developing multi > > I am hearing about SEDA for the first time and it says it is a 'Staged > Event Driven Approach' and gives the details of how staging can help > in the Event Driven Approach. Now, we have quickly moved to a 'staged' > event driven from event driven. Pretty new concept, at least in the > Python world. How is twisted coming into picture here? > > But yeah, this discussion is pointing out that that a single focussed > topic, say if you take twisted and designing a server using it might > be helpful. > > If focussing on Benchmarks, then it might be a topic in itself. > > > -- > Senthil > > BOFH excuse #402: > > Secretary sent chain letter to all 5000 employees. > _______________________________________________ > BangPypers mailing list > BangPypers@python.org > http://mail.python.org/mailman/listinfo/bangpypers > _______________________________________________ BangPypers mailing list BangPypers@python.org http://mail.python.org/mailman/listinfo/bangpypers
