The best ontological definition of document, I was able to find is in “Documentality: Why It Is Necessary to Leave Traces (Commonalities)“, by Maurizio Ferraris, http://www.amazon.com/Documentality-Necessary-Leave-Traces-Commonalities/dp/0823249697 that is the translation of “Documentalità: perché é necessario lasciar tracce” by Maurizio Ferraris http://www.amazon.it/Documentalit%C3%A0-Perch%C3%A9-necessario-lasciar-tracce/dp/8842091065/ref=sr_1_4?s=books&ie=UTF8&qid=1355064359&sr=1-4 .
Documents are on the top of a four layered pyramid of signs.
On the left hand we have the escalation of objects towards documents and on the right hand we have the escalation of signs towards the documents. At the apex documents and objects merge: a banknote, a passport, are at the same time objects and documents. Marcel Duchamps “Still life with chair” is an object and a document: http://www.verkstad.com/images/art/duchamp/bicycle.jpg
At the bottom of the documental pyramid we have the trace. It is a sign that has been generated by events, with no meaning, intention or whatsoever purpose. In IT the only example of trace that comes to my mind is the trace left by deleted data on a memory support.
At the next level we have the recording, where the trace is generated on a medium that is (ontologically) designed for keeping the trace over the time: like the recordings of a camera or a microphone left switched on, or like something that I passively memorize in my brain. Recordings have to be accessible at least by one person. They are not created with a specific meaning or purpose, they simply exist on a medium that is designed with the purpose to record facts. The difference between trace and recording is the functional nature of the medium: you have on a wall the trace of a flooding; on a hydrometer the same trace is a recording. In the digital environment, recordings are all digital data generated by a given IT system, without the aim to be seen by anyone else then the system administrator.
At the next level we have inscriptions, where the recording is ontologically accessible by more then one person: inscriptions are recordings that are made intentionally for the purpose to be read by more then one person. The intention of the author of the inscription is both: aimed at making the inscription and at making it accessible. An inscription has no inherent semantic aim: the author of the inscription does not want anything more, then to leave a sign and make it available to at least one more person. The intention to leave a sign differentiates the inscriptions from the recordings. Inscriptions are in the digital environment all organized data, that are kept together to provide information not only to the system administrator, but also to other authorized persons (like auditors, experts, etc.).
Documents are the inscriptions that the author has generated in order to convey a specific semantic message. If there is the further aim to have a socially recognized effect, we speak of legal documents (or strong documents); if the aim of the document is simply to convey a specific information to at least one additional person (think of an encrypted love message), with no whatsoever socially accepted consequence, we speak simply of documents (or weak documents). The difference between strong and weak documents are of huge social importance: in fact strong documents may have to be generated in a formal context, may have to use a language that is formally accepted by the social context (in order to avoid ambiguity), may need to have an author that is defined or definable, may need to have a recipient that is defined or definable, may need to be fulfill specific formal requirements. But most importantly of all, the meaning of a strong document is strongly determined by the social context, and the individual intention may become just one of the factors that determinate its final meaning.