At the New Zealand Electronic Text Centre (http://www.nzetc.org/) we make use of a MADS authority file for the names appearing in our collection of texts (full text and images of dead tree works). We have some issues with the schema, and would appreciate feedback on our usage - it may be that another schema might better suit our needs, or we may not understand some aspects of MADS. The following description may also prove interesting in light of the recent discussion on alternate personal names in MODS. We use our authority file to keep track of the names of every person, organisation, place, title and ship (we have a large war history corpus) mentioned in the works we have digitised.[1] So far each entry is a record not of a bibliographic entity but a 'real' one - an entry is for a person, not for a name.[2] We use the authority part of an entry to give the fullest version of a person's name we know, preferably in its native language and script. The variants represent other names, abbreviated forms, the name in other languages and scripts, etc, as (I hope) is usual. These entries are each given a unique ID[3], and this ID is used in the electronic texts. We use TEI (http://www.tei-c.org/), and mark up names as, for example, <name key="name-207379" type="person">John Beaglehole</name>. The text within the markup is whatever appears in the original work, while the key is the ID of the MADS entry - the two are entirely independent (except of course in that they both refer to the same 'real' entity). In order to mark up the electronic texts, we have a set of scripts which match a provided string against the authority file, checking authoritative and variant names for a match, doing its best to cope with initials, ordering, and the like. The ID is of central importance, since it is what is used in all of the relationships that underpin the site. A text (which has an ID) is written by one or more authors (each with an ID), with a publisher and place of publication which are also represented by IDs. The actual names associated with a record are of secondary importance, in that we can change them as we wish without any of the established relationships being affected. We can make corrections, add new variants, replace the authoritative name with a different one, etc. In addition to marking up places where a name occurs (or an entity is referenced[4]), our site will often show a name where all we have is a MADS ID. So, far example, the page http://www.nzetc.org/tm/scholarly/name-207379.html is titled "John Cawte Beaglehole", with that name being the authoritative version of his name in the MADS file. On occasion such a heading will include two names, when the authoritative name is not written using the Latin script (eg, http://www.nzetc.org/tm/scholarly/name-025840.html). This means that a link from a document which uses one name may lead to a page which is titled with a completely different name (eg, Mark Twain to Samuel Clemens). Obviously it would be useful to provide (on pages detailing a particular 'real' entity) all of the variant names. Unfortunately, it does not seem possible to reflect all of the relationships between variant/related names using MADS. For example, it would be nice to specify that a variant name is a maiden name, or that a translation is in fact a translation of a variant, not of the authoritative form. In the future, we would also like to model genealogical relationships between people, but that is well outside the scope of MADS. I hope this description has not been confusing! Jamie [1] That is the goal; in reality many names are included, and many instances of a name we do have are not marked up in the electronic document of a work. [2] We have not (yet) encountered a situation where a single bibliographic entity maps to more than one 'real' entity. We're not sure what we will do if and when that occurs. [3] We adopted MADS when the latest draft did not have an ID attribute defined for the mads element, which caused some confusion as to how others were referencing their own records. [4] For example, in marking up some bibliographies which use "--" to indicate the same author as the previous entry, we mark this up as <rs key="name-207379" type="person">--</rs>.