[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [regrep] [Urgent Issue] Should reference attributes in rimclasses be elements of type ObjectRef
The use cases for Object References in general are: * Static reference from one object to another. Example: sourceObject in an Association statically references a registryObject by its id * Dynamic reference from one object to another. Example: sourceObject in an Association dynamically references a registryObject that is the singleton result of a QueryInvocation defined within the dynamic object reference * Remote object reference from a object in local registry to an object in a remote registry. Example: sourceObject in an Association statically or dynamically references a remote object in another registry The nice thing about the proposed change is that there are no caveats in the spec that describe what a server should do based on the value of a reference attribute. Instead it would describe what a server should do for any element of type ObjectRef based upon the specific sub-type of ObjectRef. For example spec could define a canonical query GetReferencedObjectQuery which is given the id of an ObjectRef. The GetReferencedObjectQuery could be defined to fetch the statically or dynamically resolved target object based upon the concrete type of the ObjectRefType. The bad thing about the proposed change is that it would require more space in database implementation and more joins in queries involving reference resolution unless implementations were to do some complex optimizations. In either case the functionality is likely to be the same for end user but in the proposed change it would mean 2X-5X more storage and 2-5 extra joins in many queries. The most typical scenario is static object references to local object for which our current schema is highly optimized. Our current schema can be used for the other two cases by defining some special handling of values for reference attributes to handle the less common dynamic and remote object reference cases. From an application stand point we could minimize impact in either option by providing the canonical GetReferencedObjectQuery with appropriate parameters and semantics defined. Oliver Newell wrote: > > Farrukh - > > I have a similar comment to Carl's - I don't have a clear picture of > the potential benefit from the end-user perspective, so it's hard to > weigh the pros/cons of the proposed change. > > Do you have an example of a concrete use case (or cases) that it > addresses, and some estimate of how often the use case crops up? > > -Oliver > > Farrukh Najmi wrote: >> Dear Colleagues, >> >> The following issue is one that needs urgent input and resolution due to >> its global impact: >> >> <http://wxforge.wx.ll.mit.edu:8080/jira/browse/REGREPTC-6> >> >> Please review, ask questions and share your thoughts on this thread. We >> need to resolve this issue by our >> next meeting. Thank you. >> -- Regards, Farrukh Najmi Web: http://www.wellfleetsoftware.com
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]