Closed gra-moore closed 12 years ago
Uh, are we using the feed URL to identify the graph? What feed URL? (Serious question, since there's more than one.)
Fair point, that's not clear and also the feed could change but the identity of the underlying 'collection' should be consistent. Maybe there should be a property on the collection item that provides the collection identifier.
Yes, exactly. For example, if you make the feed URL be the graph URL then you lose the ability to syndicate feeds.
I think a property to identify the feed is the right thing to do.
We have a property, and put it in the collection, fragment and snapshot feeds. We call it collection-id (decide on correct capitalization later.)
Just doing some work on the B* implementation and thinking about this some more. In the spec at the moment we say nothing about mapping a feed to a specific graph or store etc on the client. So i wonder if we really need this collection-id thing at all? A client is configured to read from a given collection and update a particular store / or graph in a store. if the spec isnt going to say how these mappings are done then why do we need this property at all?
Please see last comment. In general i feel that letting the server define anything about where and how the client stores data is wrong.
We kill this, and put it into client config. Tighten text in 5.2.4 to refer to a specific graph G, but the client chooses G.
Done.
Now that we are using the feed url to identify the graph and in turn that to slice up data from different sources (and not as before using the subject locators of occurrences and associations) then this extra extension element seems redundant. Any objections?