(this information is cited and modified from the European NEEO project. Network of European Economists Online (NEEO): project information see http://www.nereus4economics.info/neeo.html. For the DAI information see specifications: http://homepages.ulb.ac.be/~bpauwels/NEEO/WP5/WP5%20Technical%20guidelines.pdf )
Building dynamic publication lists per author requires that these authors are unambiguously identified. This is best done through a unique identifier that is assigned to each author of a work. Such an author identifier is called a DAI (Digital Author Identifier).
A DAI can be assigned to authors on a national level (like in the Netherlands where each author gets a unique identifier in the METIS system), or on an institutional level. It is the sole responsibility of each IR to ensure that an author can be identified through a DAI and that each assigned DAI is unique within an IR.
Format of a DAI
Every IR can deliver its DAI's in the format it wants, as long as the authoritive party that acts as a Registration Agency can be recognised in the scheme. However it is recommended to use the International Standard for Name Identification (ISNI). This is a standard in development, No Registration Agencies set-up so far. The project finishes in 2009. The DAI numbers in the Netherlands are ISNI compliant due to involvement via OCLC.
http://www.collectionscanada.gc.ca/iso/tc46sc9/docs/sc9n429.pdf number. All DAI's MUST be globally unique. This is accomplished by combining the DAI with its authority (value of the authority attribute of the identifier element) or by making the DAI a complete URI that is unique. Some examples of valid encodings of a DAI:
Persistence of a DAI
DAI's should be Persistent Identifiers: a change of DAI for an author could effectively result in incoherent results for service providers worldwide and publication lists could become incomplete. For example, part of a publication list would be allocated to DAI X, another part to DAI Y, both DAI's referring to the same author. Statistics on downloads of publications per author would also become incorrect. If an institution needs to change the DAI's of its authors, for whatever reason, a complete re-harvest of the IR should be operated by all service providers and link resolvers on a global scale, in order, for example, to get the publication lists right again. Errors in usage statistics services would probably be irrecoverable. The advice is clearly that DAI's shouldn't change, once they are assigned to authors.