DOIs can identify digital objects but can represent other things. The DOI handbook says "A DOI name can be used to identify any referent involved in an intellectual property transaction". (http://www.doi.org/handbook_2000/intro.html#1.4.2)
I tend to think of it as a 'digital' object identifier, rather than a 'digital object' identifier :)
They are most used (afaik) for journal articles, but also used regularly for ebooks and conferences etc.
Following an announcement last year http://eidr.org/newspress/leading-entertainment-companies-create-registry/ we can expect to see DOIs become common for identifying movies, tv programmes etc. - although I'm not sure if/when these might start to enter the library data chain.
In terms of what the DOI 'does', the DOI handbook says "the simplest action that a user can perform using a DOI name is to locate the entity that it identifies. In this respect, a DOI name may look superficially like a URL. However, the technology which underlies the DOI System facilitates much more complex applications than simple location"
The other common use of the DOI is to find a description (metadata) of the thing identified using the 'CrossRef' service - some further information on how you can use this at (and linked from) http://www.crossref.org/CrossTech/2011/04/
I thought there was now a recommendation that rather than using a bare DOI, you should always use a http resolvable version (that is the DOI as part of a URL using the 'resolver' http://dx.doi.org - e.g. use http://dx.doi.org/10.1126/science.1157784 not just DOI: 10.1126/science.1157784). Unfortunately I can't find the information about this right now.
Owen Stephens Consulting
Email: [log in to unmask]
Telephone: 0121 288 6936