>>The recent tag reuse changes appear to be great, but seem to have the >>disadvantage of breaking existing tags (for example, the Foo tag in the >>examples, as well as two of the tags from the admin webapp). Although >>none of these tags was compliant with the specification (both had an >>iccorect state handling), this still raises the problem that the change >>will likely break existing web applications, some of which may be hard / >>impossible to fix. > > This is one of the reasons that tag-pooling is disabled by default in the > 3.3 Jasper. The other reason was that in some cases it hurt a lot. I > haven't looked at the Jasper2 code, but from the discussion it doesn't > syncronize as much as in 3.3, so that may not apply to Jasper2.
Since the testing was made with Jasper 1, I am not convinced that the testing was accurate because of the impact of the try/catch nesting performance bug. Tag BodyContent buffer reuse (esp with the old 8K buffers) likely has a bigger performance impact than tag instance reuse, but it would probably also have been offset by the try/catch bug. I mostly fixed the bugs the 4.1.4 build had, but 4.1.5 will be a nice candidate to make sure that reuse does indeed help performance. I hope I'll be able to add "big time" to the previous sentence :) Remy -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
