On Tue, 29 Mar 2005 11:04:20 -0300, Edgar Poce <[EMAIL PROTECTED]> wrote:
> Hi stefan
> 
> >>1 - referencial integrity
> >
> >
> > referential integrity is taken care of by the repository. it's not a
> > task of the persistence manager.
> >
> I think it's not a problem in controlled deployments where the developer
> or a support team is in charge, but there are many situations where it's
> not the case, and if the repository turns unusable it's a problem both
> for the client and the company that sold the product. Given the fact
> that if the jvm is killed the repository might turn inconsistent I'd
> rather use a persistence storage that checks referencial integrity.
> 
> >>3 - not duplicated data
> >
> >
> > again, that's taken care of by the repository.
> With high load and a networked deployment (usually together) unnecesary
> bytes would bother the network. There are cases when this situation can
> not be easily fixed, but in this case the PM can easily chose to ignore
> some store calls, e.g. all NodeReferences and only the primaryType
> PropertyState.  Actually, as Tobias pointed, the versioning part choses
> not to pass the primary type as a PropertyState in the ChangeLog, I
> guess the same can be done easily in the PM.

i disagree. in fact that's exactly what a PM is *not* supposed to do. 
jackrabbit was not designed to accomodate *smart* PMs. 

cheers
stefan

> 
> regards
> edgar
> 
> >
> >
> > cheers
> > stefan
> >
>

Reply via email to