> An interesting option for a JDO implementation is to use a RESTful > service like Amazon S3 or memcached for the back end. Some obvious > issues are transactions, object identity, and query. But having an > object front end for these would have clear productivity benefits > compared to a user writing the entire thing by hand (tools are pretty > simple for accessing data from these services).
If someone is interested, they could easily write a DataNucleus plugin to do that. 1. Would take maybe 2 days to do basic persistence (insert/update/delete/fetch/locate) - based on previous experiences with LDAP, XML, Excel, JSON datastores etc we support. 2. JDOQL (and JPQL) could be provided easily using in-memory queries (that DataNucleus provides), and then extended to make use of whatever Amazon query language is available at some later stage. -- Andy (DataNucleus - http://www.datanucleus.org)
