Carsten Ziegeler wrote:
Sylvain Wallez wrote:
However, I still consider the new implementation of SourceResolverImpl.release() in Excalibur (dated 15 dec. and not currently in Cocoon) to be a performance burden since in most cases all component-related operations it performs *will be useless*.
Carsten, what do you think ? Should we move this to avalon-dev or can you wear you Avalon hat here ?
I'm glad that you wrote hat and not hate... (just a hint of a joke).
Wow, this kind of typo must absolutely be avoided ;-)
The implementation currently used in Cocoon is wrong. The new version i Avalon is IMHO cleaner and correct, because a factory is responsibleThe performance problem is that among all implementations of Source that I know of (URLSource, FileSource, SlideSource, BlobSource, XMLDBSource, SitemapSource and CVSSource), only one actually needs to be disposed (SitemapSource).
for creating a source, so it must be the responsibility of exactly this
factory to pool/destroy/recylce the source as well. It cannot be
the source resolver doing the job.
Where do you see the performance problems?
So having a isSelectable()-select()-release() cycle to call a factory.release() that does nothing most of the time seems CPU waste to me.
Now I agree that theoretically releasing is the factory's job. So, in order to lower my performance fears, what about removing the isSelectable() call in SourceResolverImpl.release() ? It would remove an unnecessary search in the ServiceSelector HashMaps, considering also that the select() that follows this isSelectable() is surrounded by a try/catch.
Sylvain
--
Sylvain Wallez Anyware Technologies
http://www.apache.org/~sylvain http://www.anyware-tech.com
{ XML, Java, Cocoon, OpenSource }*{ Training, Consulting, Projects }
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, email: [EMAIL PROTECTED]
