Jeni's Musings

I think either “alternate” or “self” would do, but they don’t really have the right semantics. I want to say “this is the canonical, context-free version of the resource: the URL you should use if you’re talking about this resource”. An “alternate” implies that all alternatives are equal (as does “owl:sameAs”). I thought about “self” but when you’re viewing an item-in-the-context-of-a-collection, then there’s an argument that the page is a view of the collection rather than the item (or at least I think the semantics are sufficiently muddled that “self” doesn’t really work).

Imagine you’re looking at a summary or abstract of a document, and the title is linked to a page containing the entire document: what “rel” do you use on that link? That’s the relation I want.

On your query-as-resource points:

Interesting idea about hashing the query parameters. I guess the main difference between that and just embedding the query in the URL is that you obfuscate the query and shorten the URL. I was really thinking about situations where people would give queries names and descriptions; since these would be dependent on the person who created them, you wouldn’t want to base the URL on just the query parameters.

Similarly for garbage-collection: if queries are resources like any other, then you use the same garbage-collection set-up as you do for other resources. For example, if they’re generated by authenticated users, then it’s up to them to manage their own space, and if you allow generation by non-authenticated users then you might give them a short shelf-life, or only allow non-authenticated users to have one query on the go at a time. Either way, though, the results of the query should be GETable by non-authenticated users.