No, not own in an IP sense, but in a property sense. There are thin clients that you can use to maintain your weblog. Some of these clients maintain copies of your posts that also reside on the server. Which one is definitive? I.e.:

Does the client or the server own your posts?

Ideally the distinction should be hidden from the user, who would prefer to think of there being only one post that manifests itself in different places.

I.e. changes to a post made in one place should sync up with the other place ASAP. (Which means hopefully immediately, or at least the next time the client can establish a connection to the server.)

The same issues already exist with IMAP mail. The client can make changes to the mail store (deleting, moving messages) while offline, and the client syncs up when it goes online. Or multiple clients can be running. It doesn't always work perfectly, but usually pretty well.

Part of what's necessary to do this is to realize that what gets communicated from the server to the client isn't necessarily posts but events. If a post is edited or deleted from somewhere else, the client needs to discover that when it syncs up. I don't believe Atom has any provision for this yet.

Posted by Pillbug on 2003-10-28

The server 'owns' the post.  The client is the device that communicates the post to the server, but until the server receives it, it doesn't exist, regardless of what exists on the client.

Just my opinion...

Posted by Chris Hansen on 2003-10-29

I rather like Pillbug's point re. events. How about: the web owns the post (i.e. it will have a persistent URI etc), the client and server merely interact with the web...

Posted by Danny on 2003-10-29

The answer might be that there is no definitive "owner" of a post. Pillbug makes a lot of good points about IMAP, but there's more to be learned from the IMAP world, namely 1) the spec is always evolving and 2) server and client apps are at various stages of implementing the features. For example, IMAP supports user-defined flags but right now only very few servers or clients support these. When I use Mulberry with an old Courier IMAP server, I'm able to use user-defined flags on my local copy of a message but the Courier IMAP server doesn't support that. Fortunately when I do a sync, the user-defined flag is maintained on the local copy even though it's not possible to use it on the server copy. Conversely there are examples of a server supporting a feature that a client doesn't support. I think what we want is a way to say that a post is the superset of all its various incarnations, where conflicting info is decided based on timestamp or something else.

I hope this make sense.


Posted by Nancy McGough on 2003-11-10

  Yes, that does make sense. The robustness of that kind of interaction between the client and the server for IMAP is also the kind of interaction I expect to see out of clients and servers that implement the AtomAPI. That kind of robustness will come from picking a document-literal form of message as opposed to an RPC based mechanism.

Posted by Joe on 2003-11-10