posted 20 years ago
I don't know what ESRI uses internally, though I am willing to bet they have built their own home-brewed database on top of the file system, to get the necessary performance they needed. They probably could/should have used an object database instead. You end up spending a LOT of time developing your own home-brewed database. I know, we did it all the time in Bell Labs.
No, I am referring to products like the object databases from FastObjects (formerly Poet) and Versant, which have been around for quite a while. FastObjects has had their JDO implementation out for a year now, Versant has still not released their JDO release. But Libelis offers a JDO interface to Versant now.
There is also a completely new vendor called ObjectDB, which is based in Israel. They have a pure-Java object database, built from the ground up to support JDO. They were hoping to have a final release available this month. I have not tried it out yet, but plan to.
I have had very good experiences with FastObjects, both the product and the company. They offer free downloads for evaluation of their product. They also provided the resources for JDOcentral during its first year launch. For just learning how to use JDO, try their j2 product.
One difference with using an object database is that they completely derive their internal schema based on your object model. So you don't have to define any OR mapping information for JDO, the JDO metadata primarily consists of listing the persistent classes and identifying the element types of collections. This requires less work on your part to get things up and running.