| 摘要 |
The directory service of the present invention extends the mechanism used to define ordinary properties to define "links," pairs of properties that represent an interobject reference with a certain meaning. Each pair is given a unique "pair ID" and consists of a "link" property and a "backlink" property, which represent the "reference" and "is referenced by" semantics, respectively. Rather than storing interobject references with the referencing and/or the referenced objects themselves and imputing the link, instead the present invention stores the link instance itself as an unnamed object in a link database table, which is separate and distinct from the data information base, and imputes the values to be reported for the relevant properties on the referencing and referenced objects from the existence of records in the link table. When a directory service server learns of the deletion of an object (either via replication or because of a deletion occurring on that server), it removes from its link table all records that refer to that object in either the link ID or backlink ID column, which is an efficient operation. This eliminates any dangling references to the deleted object. Although this implicitly affects the values of the corresponding link and backlink properties on other objects in the directory (i.e., those named in the backlink or link properties of the object being deleted), the changes to those other objects do not need to be noted or replicated in any way, since the replication of the deletion of the original object will cause all other directory service servers to perform the same cleanup, thereby independently removing the same link instances. This allows the directory service to maintain referential integrity in a distributed manner, with minimal replication overhead.
|
