Re: [hibernate-dev] SessionFactory building, 5.0-style

I assume no responses means everyone is ok with this change. Hopefully I will start on those changes next week; for certain I would like to get it done before the meeting On Aug 23, 2012 8:28 AM, "Steve Ebersole" <steve(a)hibernate.org <mailto:steve@hibernate.org>> wrote: The initial design I had for building a SessionFactory using the new metamodel was 3 steps: 1) build ServiceRegistry 2) build Metadata 3) build SessionFactory using both ServiceRegistry and Metadata This changed a little as actually implemented in the metamodel branch: 1) build "boot strap" service registry 2) build basic service registry 3) build MetadataSources 4) build Metadata 5) build SessionFactory using both basic service registry and Metadata I would like to change this slightly based on JPA 2.1 work and integrating that with containers (mainly through planning with Scott for JBoss AS). <background> Essentially Scott and I made a change proposal to JPA EG for how managed EMF bootstrapping happens to better account for stuff in the container's environment not being available until certain times. It is a typical "hole in the interaction of specs" deal. Long story short, we want to make boot strapping of an EMF into 2 distinct phases. Whether or not that gets accepted/approved, we will implement this approach for Hibernate EMF bootstrapping and JBoss AS will leverage it. The 2 phases are meant to account for container resources not being available or the need to delay classloading. </background> The changes I propose would just be ordering: -- phase 1 -- 1) build "boot strap" service registry 2) build MetadataSources -- phase 2 -- 3) build basic service registry 4) build Metadata 5) build SessionFactory using both basic service registry and Metadata -- steve(a)hibernate.org <mailto:steve@hibernate.org> http://hibernate.org