Editing Directory service (section)

== Comparison with relational databases ==
Several things distinguish a directory service from a [[relational database]]. Data can be made redundant if it aids performance (e.g. by repeating values through rows in a table instead of relating them to the contents of a different table through a key, which technique is called [[denormalization]]; another technique could be the utilization of [[Database replication|replicas]] for increasing actual throughput).<ref>{{Cite web |title=When and How You Should Denormalize a Relational Database |url=https://rubygarage.org/blog/database-denormalization-with-examples |access-date=2023-04-30 |website=rubygarage.org |language=en-US}}</ref>

Directory schemas are object classes, attributes, name bindings and knowledge (namespaces) where an object class has:
* ''Must'' - attributes that each instances must have
* ''May'' - attributes which can be defined for an instance but can be omitted, with the absence similar to NULL in a relational database

Attributes are sometimes multi-valued, allowing multiple naming attributes at one level (such as machine type and serial number [[concatenation]], or multiple phone numbers for "work phone"). Attributes and object classes are usually standardized throughout the industry; for example, [[X.500]] attributes and classes are often formally registered with the [[Internet Assigned Numbers Authority|IANA]] for their object ID.{{citation needed|date=June 2019}} Therefore, directory applications try to reuse standard classes and attributes to maximize the benefit of existing directory-server software.

Object instances are slotted into namespaces; each object class [[Inheritance (object-oriented programming)|inherits]] from its parent object class (and ultimately from the root of the [[hierarchy]]), adding attributes to the must-may list. Directory services are often central to the [[Computer security|security]] design of an IT system and have a correspondingly-fine granularity of access control.