On Sat, May 28, 2011 at 7:44 PM, Greg Wilkins <gregw_at_intalio.com> wrote:
> Rajiv,
>
> One area that badly needs attention for cloud support in the servlet
> spec is the HttpSession distribution semantics.
>
+1
I had proposed for JSR-315 to add a "distributed" directive for sessions,
similar to the existing "distributed" container, so the requirement to have
"all the requests to be served by the same instance at any time" can be
relaxed.
> We still have massive ambiguities in the spec about how and when a
> session is distributed/persisted/serialized.
> Eg. some clustering solution treat values set in the session as pass
> by reference, so that changes made to the objects after the
> setAttribute are reflected in the persisted/distributed state. Others
> treat the setAttribute call as pass by value and take a copy of the
> state at that time so that changes made after the set are not
> reflected in any migrated/restored sessions (but are probably seen by
> other requests in the same container).
>
> I think the problem that we have is that all interpretations of the
> spec probably have some merit, so we are not going to be able to firm
> up the current semantics without breaking one clustering solution or
> another.
>
> 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 ;)
> Is that the kind of improvement you see we should be considering ?
>
> cheers
>
>
>
>
> On 28 May 2011 15:46, Rajiv Mordani <rajiv.mordani_at_oracle.com> wrote:
> > The main theme for Java EE 7 as specified in the platform spec JSR and
> also
> > as a main requirement for Servlet 3.1 is enabling the platform for use in
> > the cloud. Linda and Bill have started a discussion on the topic over on
> the
> > javaee-spec project at the platform level. Some of these requirements
> will
> > come down to the Servlet spec eventually. I would like to encourage
> people
> > in this expert group to read the proposal at
> >
> >
> http://java.net/projects/javaee-spec/lists/jsr342-experts/archive/2011-05/message/11
> > and either participate in the users mailing list for the javaee-spec as
> it
> > is following the same model for transparency of the JSR as this one, or
> if
> > you rather not join the discussion in the users mailing list for some
> reason
> > but still have some feedback on the proposal, you can send the feedback
> to
> > me which I can pass on to Bill and Linda.
> >
> > Thanks
> >
> > - Rajiv
> >
> >
>