> The intent of the "increment" strategy is to allow use of a database > that the user has no control over (for example, the DBA refuses to add > new tables). It's not an optimal strategy but a useful one. The JDO > implementation can cache the largest key used and mitigate database > access for each insert. > > But without changing the specification, I don't think it's ok to > require another table in order to implement "increment". That's what > "sequence" is for. > > Was there a recent change in DataNucleus that now "increment" is > implemented using an internal "sequence" strategy?
DataNucleus simply changed to make use of the "autoCreate" flags in imposing whether it was allowed to create *any* schema components. It has ALWAYS used a table for "increment", that is not a change. "sequence" has always been to use a datastore sequence also. -- Andy DataNucleus (http://www.datanucleus.org)
