Identity

An identity is a label, usually unique, that identifies some artefact.

Characteristics of good identities

 * Uniqueness
 * Uniqueness is context dependent, but to be useful an identity must be unique within its defined context. Consider, for example, the ISBN identity assigned to books. This is a series of digits that is only guaranteed to be unique within the context of being an ISBN (any given sequence of digits used as an ISBN could be used to identity something other than a book but would not be an ISBN).


 * Stability
 * Identities should be stable, not changing over time. Once assigned an identity must always refer to just one thing and that identity must never be reassigned.

Poor identities
Any identity which is not a good identity is usually, by definition, a poor identity. The following are broad indicators of poor identities.


 * Variation over time
 * An identity that varies over time is a bad choice. For example, it is common to refer to 'the latest version' of something. This is not, however, an identity. It is a specification.


 * 'latest version' varies with time. The latest version today may not be the same artefact today as tomorrow.


 * Ambiguity
 * In everyday speech we use pronouns to refer to things, 'his laptop', 'her USB drive', 'my monitor'. Not only are these references 'variable over time', they are also context dependent and worse the context is time dependent ('his' and 'her' depend on the context of the specific conversation).


 * This may seem irrelevant to identification, but consider the 'identity' Marketing Department Laptop. Is there more than one marketing department in the organisation? If so, to which marketing department does this refer? Does the marketing department have more than one laptop? How is laptop being defined? Although in everyday speech we might refer to 'fixing a problem with the Marketing Department Laptop' there are too many ambiguities to make Marketing Department Laptop a good identity.


 * Location dependent
 * Identities that tie artefacts to locations tend to be poor choices (the exception being identities of locations). Suppose we chose to identify desktop computers using a simple scheme such as where  identifies the site to which the computer was deployed. This seems to be a good idea. It allows us to easily see where a desktop computer is deployed and helps with asset management. The problem comes when computers are redistributed. When a computer is relocated we have two choices, either; change its identity to match the new site (violating the idea that the identity is persistently applied to the artefact—the computer in this case), or keep the identity and lose the reliability of the information embedded in the identity.


 * Some challenge the first problem (that changing the identity to match the site is an issue). This is a consequence of misunderstanding the importance of identity persistence. If we allow the computer to be relabelled with a new identity we introduce the problem of tracing the artefact throughout its history. Suppose the computer was originally labelled as LON1234, identifying it as located in the London office, and later it was relocated (and relabelled) as CAR1234 and hence at the Cardiff office. Later, looking through old change records I encounter a reference to LON1234 but I cannot find that artefact anywhere (because it is now called CAR1234.


 * We could maintain a trace by recording that LON1234 was relabelled CAR1234 and so when LON1234 is encountered in a change record we could look this up and see that it is now labelled CAR1234, but this adds a significant overhead to the identification process. Not only must all relabelling events be carefully recorded (and any identity on the artefact itself changed) but we must be careful to never reassign that label and any lookup must now trace through all of the subsequent relabelling events. While these issues are not insurmountable they do indicate that embedding location information in identities leads to poorer identities. (As a side note, location information is better recorded as either an attribute of the artefact or as a relationship between the location and the artefact in a CMDB.)