[Chandler-dev] Server-side backups theoretically desirable?

[Grant Baillie]

On 24 Aug, 2006, at 10:12, Andi Vajda wrote:

>
> On Thu, 24 Aug 2006, Brian Moseley wrote:
>
>> we've talked about version numbers in regard to making syncing more
>> efficient. this is another way they'd be useful. i think grant's
>> suggestion of examining the last-modified date in combination with  
>> the
>> etag is a more interoperable way of doing the same thing, though.  
>> yea,
>> time is fuzzy, but for the use we're talking about, etag+last- 
>> modified
>> would be sufficient almost every time, i'd think.
>
> If the last-modified values we're comparing are all issued by the  
> same server then it should work as well as a version number. It's  
> comparing client time vs server time that is flaky.

That's one reason for flakiness. Another is that Last-Modified only  
gives you 1 second precision, which isn't really enough to resolve  
collisions from multiple clients writing to a resource.

--Grant