On 10/31/05, Benoit Masson <[EMAIL PROTECTED]> wrote: > > Hi you all, we're about to start a project on the web. We are quite > seduced by the turbogears framework. But for now turbogears is still > in developpement. Nevertheless it is based on mature framework > cherrypy which seems to be quite robust. Other framework are also > reaching 1.0 version... So my questions are: > - if ou project is to ba online in 3 month for testing, do you think: > I should start development with 0.9(indentity, catwalk and so...) ?
0.9, as of this very moment, is still a moving target as far as its API is concerned. Hopefully, it won't be a very long wait before 0.9 is out, though. At that point, that would certainly be the API to write to. > - our project should have quite some hits/hour do you think > turbogears, associated with apache as frontend will fits that need ? CherryPy/TurboGears can process a bunch of hits. It really boils down to what the application is doing. There are a myriad of ways to improve performance if you do find a hotspot in your application. > - What about load balancing ? Willl we be able to use cluster fot the > cherrypy part ? Separating the database is a basic move, but what > about turbogears sharing sessions between server ? or should we look > LVS cluster instead so the framework doesn't even know it is on a > cluster ? If you can avoid sessions (through judicious use of cookies, form variables, iframes, etc on the client and database model objects on the server), your load balancing picture becomes much easier. HTTP is inherently stateless and the more you can do to preserve that, the easier you can scale. Gerhard had good suggestions for the sessions, too. Kevin
