Hi, I have followed the discussions on benchmarks and I have some further comments:
Be careful with general statements! A really descriptive statistic would be more useful. For example I have observed that for big flat XML structures Xalan is twice faster than Saxon. On the other hand for deep structures Saxon is faster. In other cases I haved observed that Xalan and Saxon are more ore less equal wih using cached stylesheets. During the evolution of these processors the lead in performance for my test cases has changed some times. E.g. 4 month ago Xalan was faster, 2 month ago Saxon gets a perfomrmance improvement and then was faster again and at the moment I have no numbers. But if you don´t use stylesheet caching for my test cases Saxon is 1.5 times faster. What about conformance? A percent value makes absolutely no sense. How do you define a value other than 100%? As a developer you must know what works and what not. Just a comment to Xerces: The most relevant design objective of Xerces 1 was performance, but not for Xerces 2. Xerrces 2 has some other most relevant design objectives. But for example Xerces 1 does not allow grammar caching. And with grammar caching for my test cases Xerces 2 runs 10 times faster than Xerces 1!. So in my opinion a real statistic describing and analyzing what is really going on would be a useful and fair solution. Best regards Thomas
