I haven't benchmarked this, but the lookups are basically a series of hashtable lookups, which makes it a pretty speedy thing altogether. On the issue of creating a name index for components, the problem I see is that the underlying contexts are very dynamic. The meaning of "foo" could easily change out from underneath us. (if there weren't the potential for conflict, we wouldn't need namespaces in the first place) I can't see any way that the required indexing (even if it were possible for all contexts) wouldn't be a much more expensive operation. I think the best solution is that if performance is a concern, always use the fully qualified name or create aliases. In the core seam code we always use the fully qualified name. Failing that, @Import and your own custom factory values could be used to optimize lookups for anything you observe to be a performance problem. So far, of all the inefficient things Seam is doing, this is fairly low down the list of things that are likely to have an impact on overall performance. That being said, we love to see profiling results on real applications. We are 100% committed to making sure Seam isn't slowing your application down. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4125587#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...