Transient attributes

Question

Cache scheme: replicated
     If our data objects has a few transient attributes (say large attributes, that we do not like to replicate until the need arise, a.k.a lazy loading), how do we code this logic in our put/get API? An example would suffice.
     I think this issue also arises in Coherence *Web, where large attributes can be declared as transient, and are stored as different entities for performance reasons as in ``split'' session model.
     Regards,
     Bulut.

Answer 1

The easiest approach is to mark the fields as     > transient, then use getter methods with lazy loading.
     > This can read those attributes from the appropriate
     > data source (which may be a cache or even a database,
     > depending on your requirements).
     Do you think Coherence would help in this case? Where would you implement this lazy loading logic? And more importantly, how would it look like? I can see two alternatives:
     i) using Coherence services... maybe using a near cache?
     ii) independent of Coherence? which definitely looks more cumbersome.
     > Coherence*Web simply queries a method which indicates
     > whether a given named attribute should be clustered
     > or not (this class is provided by the application
     > developer). Also, note that Coherence*Web does not
     > lazily load non-clustered attributes; it actually
     > determines whether to make them visible to other
     > cluster member.
     Basically, my question is how Coherence implements handling of transient attributes, if we need a local transient attribute at a remote site.