On 12/7/06 4:11 AM, "Joe Gregorio" <[EMAIL PROTECTED]> wrote:

>>> they don't solve the lost update problem,
>> neither does Etag
> 
> You must be unfamiliar with this:
> 
>  http://www.w3.org/1999/04/Editing/

Ah, you're right. Etags are useful for the lost update problem ... but
that's a different problem to the syncing problem. Which is to say the lost
update problem is a red herring to addressing the syncing problem, and I
regret taking that bait :-P

So, let me try again, this time thinking straight ...

You said:

> [app:modified] don't solve the lost update problem,

Doesn't need to, that's a different problem, with general consensus agreeing
that Etags are useful when GET/PUT entry resources.

Lets not get the various operations mixed up:

1) discover which entries have been modified (this one, that one)
2) retrieve the modified entries (looks like this, looks like that)
3) check for edit conflicts before doing a big upload
4) upload changes without stomping someone else's edits

For these operations, do this:
1) look at app:modified in the collection feed,
   paging until app:modified is what you expect
2) GET (doesn't have to be conditional, you're expecting changes)
3) HEAD, compare Etag with the Etag from (2)
4) PUT with If-Match

Hmmm ... your proposal *would* work (meaning we'd know when to stop paging)
IFF the collection feed is sorted by app:modified (even if app:modified
wasn't in the feed, and even if that particular label wasn't used for the
concept). You'd still need to define what [your term for app:modified] means
though.

e.

Reply via email to