Grouper Developers Forum

["Jessica Bibbee" <>]

Hibernate Call
December 19, 2006

*Attendees*
Tom Barton, U. Chicago (chair, Grouper)
Blair Christensen, U. Chicago
Lynn McRae, Stanford U. (chair, Signet)
Dave Donnelly, Stanford U.
Mike Douglass, RPI
Albert Wu, UCLA
Steve Olshansky, Internet2
Jessica Bibbee, Internet2 (scribe)

*Discussion*
This call was formed after the joint Signet/Grouper Fall 2006 Internet2 Member Meeting, where there was much interest in further discussing the benefits and challenges of using Hibernate. Signet and Grouper, though separate, are both projects of Internet2 and strive to have as much in common as possible.

{Dave} is experienced with DAO, and less so Hibernate, and the question arose whether Hibernate was the right choice. Since the last Signet discussion, he and {Lynn} discovered that they might have been asking too much of Hibernate. There were many lists, sets of data for related things (e.g., subject had one set for proxies, one set for assignments, etc.)

{Mike} has experience using Hibernate, including the latest version 3.2.  He likes Hibernate; even though it is not documented well, it does do what it is supposed to do. Once he realized that he could not treat a calendar as a web application, he could accept Hibernate with its limitations. It is typically the case that a web application will fetch things in the action phase and expose them in the render phase – contrary to Hibernate, where things that are fetched do not have a live session. He recommends lightweight action at application initialization. {Albert} suggested flushing the sessions when dropping sessions.

Currently, {Blair and Dave} have worked with Hibernate 2; {Mike} agreed that there were many problems with v2, and suggested that they seriously consider upgrading to v3.2. That people switch databases to accommodate the product is not the best option; rather, it is best to work the product around the systems they are already using. Neither Signet or Grouper have been or will be locked into using one database.

{Albert} said Hibernate was created application-server module in mind, and was not meant for web services. There are different gradients of fetching - not just what is fetched, but how much is fetched. Blair is working to remove lazy fetching within Grouper. Lazy fetching will decrease efficiency in a large hierarchy.

{Lynn} suggested developing a DAO layer (Spring), look for what can be persisted in the DAO layer, and then think about Hibernate.{Dave} said that in the process of backing out of relying on Hibernate to do so much, they are moving to a more granular approach.

{Dave and Lynn} were undecided as to whether or not Hibernate 3.2 would make it into the next release of Signet. {Dave} asked if anyone had looked at the new Java Persistence from SUN, but no one attending had. {Mike} emphasized the importance of using a persistence engine, saying it removed half of the maintainable code.

The discussion was useful to all on the call, who felt much more comfortable about using [the next version of] Hibernate with Signet and Grouper. {Blair} said it reinforced the direction he was heading, which is to clean up the code and incorporate Spring. {Dave} looks forward to a less complicated use of Hibernate. The three layers will be: 1) core business, 2) Spring, then 3) Hibernate.

Discussion of Hibernate and its use alongside Signet/Grouper will continue on the respective Signet/Grouper-dev mailing lists and WG conference call.

-- 
Jessica Bibbee, Technical Analyst
Internet2

mobile: +1-734-255-6644

Wishing you an inspired and innovative 2007.
http://www.internet2.edu/greetings/newyears2007/