I make a lot of claims here which I'm still trying
to convince myself of. If anyone can analytically
set me straight on them, I'd appreciate it.
Chip Wilson <[EMAIL PROTECTED]> wrote:
>
>It doesn't matter how my fine-grained objects are
>implemented, as Entity Beans or Java Classes,
>there will still be a mismatch between my object
>model and my domain model.
I disagree that "it doesn't matter how fine-grained...
there will still be a mismatch." Direct mappings are
the easiest O-R mappings to specify. The smaller the
entities, the more likely they'll by chance map directly
to a tuple. Obviously entities should be designed from
the application's vantage point, but all things being
equal, smaller seems easier for vertical integration to
me.
>ACID has to do with transactions, not mapping.
Indeed I was taught that. In the transactional world,
isolation implies limiting a change's visibility/impact/scope.
In practice, isolation favors finer grained entities, right?
O-R mapping also favors finer grained entities, right?
I don't think that's coincidence, do you? Fine
granularity seems like a good entity design criterium. The
opposite extreme is object-graph entities, perhaps even with
entities reachable from other entities (entity graphs). I
wouldn't want to tackle O-R mapping for object graphs.
No thanks. I'd prefer small entities at
design, (re-)deploy, and run times. I also suspect small entities
are more ammenable to transaction atomicity when compared with
object graphs. If so, it's probably not coincidence that
small entities be both easier to O-R map and easier to
achieve reliable atomicity.
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
