James Snell wrote:

>John Panzer wrote:
>>>[snip]
>>> The challenge is knowing how long to keep the x:deleted-entry things
>>> around in the feed.
>>
>> Perhaps it's sufficient in many cases for the feed to indicate to
>> consumers how long it keeps deleted-entry items around.  Then a consumer
>> who checks back, finds that they may have missed a tombstone, has the
>> option of doing a complete sync by paging back through the feed history.
>>
>
>One idea I had was that a feed would only contain deleted-entry elements
>for entries that fall within the same atom:updated date range as the
>atom:entry's in the feed.  That is, if the deleted entry *would* have
>appeared in a particular feed page based on it's app:edited or
>atom:updated value, then there would be a deleted-entry element.  The
>downside of this would be that deleted-entry elements would be spread
>out across multiple feed pages.  I'm not sure if that's a really bad
>thing or not.

That actually seems like a good thing, because it means that the deleted items 
will be in the "right" place in the feed. Anyone who wants the complete update 
will need to read all of those pages eventually. For SSE (spec is on 
http://msdn.microsoft.com/xml/rss/sse), we've been thinking that a sync client 
would use ETag headers on a GET to the feed as a way to ask the server for all 
of the changes since the last sync, but I'm not sure how well that would mix 
with paging.

Regarding John's comment about about missing the tombstones--we have a 
mechanism for this in SSE. You can have an optional feed-level sx:sharing 
element with an "expires" attribute that gives the client guidance about how 
often to sync.

Steven

Reply via email to