I know that there has been some discussion here and on rest-discuss (esp. [1]) about relating existing resources on the web, and there seems to be no real consensus view about mechanisms for "creating" those relationships and "deleting" those relationships. Obviously, we have links, but how you "assert" a relationship, say, by way of AtomPub is unclear. My instinct is to simply do the GET/PUT operation on an atom:entry, adding a link. But since a link is not a resource, I don't have a good way to make any assertions about the link (other than it's @rel value) or to "delete" the link. Of course I can do another GET/PUT this time excluding the link, but that's potentially difficult to manage on the back end.
The use case is a large image collection managed in an Atom store (atompub for all operations). We need to record information about each image: bibliographic source data, country of creation, artist, etc. and when possible, we want that information to come from a standard thesaurus: Getty Geographic Name, Getty Union List of Artist names, Library of Congress Bib Record, etc. Let's say I can get an Atom representation of any entry in one of those sources (each w/ a "self" link). So I create links on each item in my image catalog. Any thoughts on how best to "manage" that relationship? Anyone else doing this sort of thing in w/ Atom? (essentially creating a relational database "on the web" where I have on table and someone else has the others). --peter keane [1] http://tech.groups.yahoo.com/group/rest-discuss/message/11383
