On 29 May 2011 20:44, Wenbo Zhu <wenbozhu_at_gmail.com> wrote:
>> But potentially we can come up with a tighter session model that works
>> well with the cloud/cluster and allow applications to opt into that
>> tighter model.
>
> I see the need for a weaker session model ;)
I mean a tighter definition of what the session distribution semantics
are. If we agree that the semantics we want include some relaxation
of concurrency constraints - then fine.... so long as it is well
defined in terms of this like granularity - by session or by
attribute. eg given a concurrent update of a session in two different
instances, will the resulting session state be the merge of those
instances with a conflict only if the same attribute was updated in
both or will 1 instance or the other be the ultimate version?
cheers