Interesting results indeed. Thanks for taking the time to test it. That will help with the performance discussion for sure.
sean On Tue, 15 Mar 2005 01:02:49 -0400, Sylvain Vieujot <[EMAIL PROTECTED]> wrote: > To get a better idea of the performance penalty involved by the > ExtensionsFilter, I did a very quick test : > > $ time wget -r http://127.0.0.1:8080/simple-webapp > > This just fetches every page of the simple-webapp. > > With the ExtensionsFilter, I get : > FINISHED --00:34:11-- > Downloaded: 588,653 bytes in 70 files > > real 0m2.914s > user 0m0.028s > sys 0m0.038s > > Then I disabled the ExtensionsFilter by commenting the code related to it, > and the by removing the ExtensionsResponseWrapper. > The application isn't functional, as no javascript & resource would be > included, but at least, it gives an idea of the performance penalty. > > Here is what I get with the ExtensionsFilter disabled : > FINISHED --00:38:00-- > Downloaded: 269,094 bytes in 50 files > > real 0m16.139s > user 0m0.025s > sys 0m0.037s > > Yes, I was quite shocked too. It's faster to get 70 files with the > ExtensionsFilter enabled that to get only 50 files without the > ExtensionsFilter ! > I did the test several times, but always with similar results. > I guess that the explanation is that the ExtensionsFilter's buffering > improves the response time, but this is only a guess. > > At least, this shows that there is no performance penalty with the > ExtensionsFilter, and it might even (against all expectations) improve the > performances for the pages. > For the resources, I'm quite sure it also improves the performances, thanks > to the client caching. > > The only case where the ExtensionsFilter could be a problem then is if you > have a page with a slow code in the middle. > It would wait for page to be completed before sending it to the client, > whereas if you have no such filter (and no compression filter either), the > user might be able to see the beginning of the page, before the server has > finished to render the page. But still, the overall page rendering time > would be very similar. > > Sylvain.
