"jason.greene(a)jboss.com" wrote : Now for an actual productive comment :) Is there any reason why the MDR has to eager allocate everything about every element of of every scanned class in memory like that? Any reason it can't just represent metadata as a series of overrides + lazily discovered / proxied calls to reflection data. Ah looking at the code, it seems this is basically what the MDR does, the problem seems to be that the API was designed for fetching specific elements, and depending on the usage, each access could be expensive (especially the reflection loader). So why does it need to scan everything, if the MDR already knows how to fetch it lazily? View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4223843#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...