On Nov 21, 2004, at 2:11 AM, Danny Ayers wrote:
In "Feed State Model" the assumption is being made that the order of the entries in the feed representation is significant (most recent at the top). I could be wrong, but I don't think that's in the spec.
Correct. I think we need to disambiguate this in any case, if there is an overall concept of a feed that's greater than the current document.
"Reconstructing Feed State", seems to me to be expecting rather a lot of the client, to remember every change and every feed head.
It doesn't have to remember every change; it just has to remember the last feed 'this' URI that it saw, so it can walk its way backwards to it.
I wouldn't have though a warning should be needed in its absence. Under what circumstances would the client actually want to reconstruct feed state? My guess is that would fall outside 80/20 applications.
Most aggregators do it today; they're just sloppy about it (because they don't have a mechanism like this). To me, the most boring aggregators are those built into browsers that don't keep state. To me, the warning is absolutely critical; it raises the guarantee you get with Atom, and differentiates it from RSS substantially.
If anything like that does need to be preserved, might it not be the receipt order of the entries? (rel="previous" on entries perhaps) I think a note to clarify sort order etc would be helpful.
Yes, it would; I don't mean to preclude the use of lexical order, just that it doesn't mean something automatically.
Is 'wholefeed' feasible/desirable? I wouldn't have thought so outside of 'control' circumstances. How do you tell the difference between wholefeed (containing all revisions of entries) and wholefeed (containing only the latest)?
Delete - yep, separate. As suggested in a recent thread, it would make most sense to have delete mean "removed from view" rather than "removed from existence". In which case it's only impact on the state would be the change of an attribute rather than any structural revision.
Whoa, heavy; I'm not sure it's interesting to talk about what's in the set 'existence'; only what's in the set 'Atom Feed'.
fyi, Henry's done quite a bit of experimentation with state and entry versioning modelling in OWL (in Bloged), though I haven't any links at hand.
Cool; if you come across them, would love to see them.
Thanks!
-- Mark Nottingham http://www.mnot.net/
