hi, I haven't commented this discussion but I've been reading it. I have to tell you that you are bouncing way out of "the point" ironically complaining that everyone else is talking about things other than the elephant in the room.
It really doesn't matter Django (unless you bring something concrete about it), it doesn't matter Massimo prior life or possible alternatives into it, it doesn't matter disqus can reach 17,000 requests per second in a very specific context where caching is allowed and the infrastructure used is absent nor if Lighthttpd can handle 10,000 unknown type of requests in an unknown hardware. I make stress testing for a living, and my reality ranges between 200 virtual users being enough to turn a powerful cluster down and me looking for more virtual machines to get enough virtual users to bring a simple server down. ..you may have something to say but you are over-elaborating your comments. A bit of reality may help you understanding things outside the enterprise corp. No, web2py doesn't have many "Corporate backing independent of lead developers", No, web2py doesn't have proven scalability based on simple metrics, No, web2py doesn't have the "largest projects". What else doesn't web2py have? It doesn't have a huge enterprise like Microsoft that owns the software that runs in most machines from the booting process until the office suite in all software layers. It's not around for more than a decade, because even dotNet didn't started from the ground when it comes to asp.net. So it's more recent, it doesn't have the time of existence required for things to happen. It doesn't have other enterprises working with it and for it including in Marketing and distribution. Oh and It doesn't have a price tag that would be prohibitive for over a half of the members of this group. What does it have? It's open source, you can use it, you can see how it works and change it. You can talk to main contributors and Massimo directly. The (expensive) phone support of .Net doesn't give you that. It doesn't depend on external tools other than web2py core itself. You can run in on any webserver, you change it's DAL to SQLAlchemy, you can run in Python, Jython, IronPython, code it with any editor with no compromise or even with just a web browser. It's open from its core with a transparent architecture that makes web2py development more like plugging python power to some smart helpers which you can manage with total flexibility. It's not a black box where you throw modules for big tasks and you shall trust them in every way (including security) so that you can make something useful of it. It's platform-independent. I run it on my Nokia N900, in my Mac, in my Linux Machines or even Windows if I thought it was a good idea to use it as a server. It's free. Now, to the "point": Yes, there is no need to handle one request by thread but that's not the strangulation point (yet). Storage is. A truly distributed, fault- tolerant and scalable storage system that provides transactions and consistency properties is in most times the point of strangulation. Keeping the web2py way of doing things you could adopt solutions that explored concurrency better. Check http://www.slideshare.net/ezmobius/erlangfactory for more about this. This is cloud computing without the buzz part. It's possible to do this keeping the same MVC programming approach but the architecture has to change a lot. You don't have this on web2py (for now) neither on .Net . Speaking of scalability on topics other than these two, storage and distributed/ concurrent task, will always be a minor talk. On Dec 3, 12:13 am, John Heenan <johnmhee...@gmail.com> wrote: > Let us be blunt about this. Web2py is out there megalomaniacally > presenting itself as all things to all people at a web framework > level. As it is now, Web2py is a very impressive achievement. > > At a purely technical level Web2py is superior to Django in a large > number of important aspects. But that is not the point. Leaving aside > the irritating Django fan-boys, Django does not present itself as all > things to all people. At the top level Django presents itself as a > commercially backed niche product suitable for simple content > presentation of small organisations (such as regional newspapers). > > If we take a straight comparison with a completely different web > framework that presents itself as all things to all people. such as > DotNet, then web2py must accept there are issues for comparison that > go way beyond pet issues found in the narrow confines of the academic > or fan-boy environments. > > Here are some simple comparison points. > > 1) Corporate backing independent of lead developers > 2) Proven scalability around simple metrics > 3) The 'largest' projects > > DotNet has been around for ten years now and has very wide > penetration. Where is Web2py going to be in ten years and what will > its penetration be? > > Frankly I cannot imagine someone with Massimo's talents as remaining > interested in Web2py, despite his deep involvement now in Web2py. > Massimo in a prior life was a talented high energy physicist, why > shouldn't he change again? Who knows, maybe a vice chancellor of the > University of Chicago trying to keep a lid on embarrassing student/ > staff scandals or fraudulent research, or lobbying for grants? > > I am not being unfair to Web2py. > > With regard to Django and Disqus, 30% of about 100 servers (about 30 > servers) just appear to run Apache + mod_wsgi. This presumably is > where Django lives. The other 70% of servers are for databases, > caching, load balancing and for other Python scripts. This information > comes fromhttp://www.slideshare.net/zeeg/djangocon-2010-scaling-disqus. > Since we can expect a lot of the requests are for the same > information, we can expect caching is very important. > > Disqus claims to be able to reach 17,000 requests for seconds or about > 570 requests per second per Apache server. Suppose each request lasts > no longer than a generous 200ms on average then Apache needs to be > able to maintain about 114 requests at once. Since Apache spawns or > maintains a thread for each request, this means that each server needs > to be able to maintain nearly 114 threads per server at once. I am not > impressed by this. > > Consider that Lighttpd can handle 10,000 requests per second ON A > SINGLE SERVER using year 2000 technology. > > John Heenan > > On Dec 2, 9:46 pm, Tom Atkins <minkto...@gmail.com> wrote: > > > > > I'm assuming John Heenan's criticisms of web2py would apply equally to > > Django? > > > Please correct me if I'm wrong. > > > Django can scale pretty well - here's a presentation showing how Disqus have > > scaled their Django app to 250 million visitors a month and a peak of 17,000 > > requests per second to Django with Apache and mod_wsgi: > > >http://nosql.mypopescu.com/post/1409154668/disqus-scaling-the-worlds-... > > > Presumably web2py could do the same?
