Re: [wsrp-coord] Groups - WSRP Coordination SC biweekly call modified

[Rich Thompson <richt2@us.ibm.com>]

Relative to your points ...

1. It might be useful to layout possibilities
for the multimedia sports scenario. Basically I think the general scenario
should include a mix of portlets where a few are from a set that were developed
together and the rest were developed independently. Most should be viewed
as having been deployed on separate servers so that Producer mediated shared
data is only partially involved.

2a. From what I understand of the Vignette
preference for managing state, allowing the Consumer to know more can be
beneficial. I think we should be careful about choosing a design where
the full model is always sent back to the Consumer. Models can grow quite
large and both the additional bandwidth and processing involved in dealing
with full models during a user interaction could have quite an impact on
performance. My current thought is to walk into areas such as these through
progressive refinements and that is why this initial question just looked
to discuss whether such transparent state is needed rather than delve all
the way into a particular design.

2b. I think the key question here is
whether events that are opaque to (but distributed by) the Consumer are
desired. Including such allows a set of portlets to have logical communication
(rather than just state based communication) even when deployment is on
separate Producers. I can see value to this, but would also like to hear
other viewpoints.

2c. Sorry that the question appears
to presume events. The key to the question is whether the additional communicated
items are piggy-backed on current responses as opposed to the other possibilities
(i.e. some form of callback to the Consumer).

2d. I haven't seen a need yet for out-of-band
communication of state, but I do think it is important we not exclude particular
implementations from using other emerging standards (or proprietary means)
to supply such when they need such capabilities.

Rich Thompson

Michael Freedman <Michael.Freedman@oracle.com> 07/29/2003 08:19 PM

To:        wsrp-coord@lists.oasis-open.org         cc:                 Subject:        Re: [wsrp-coord] Groups - WSRP Coordination SC biweekly call modified

Oops.  I forgot to send out a query seeing if
we can reschedule these 
for 9am ... as I have a standing 8am internal meeting now I can't make
this time.  Any chance we can reschedule future events for some other
time on Wednesday?  If by luck folks can postpone tomorrows meeting
until 9am let me know as I will check my e-mail in the morning before 
heading into my 8am.

As I am unlikely to be on tomorrows call, here are some thoughts on the
presentation/questions below:

1. Is someone going to flesh out the MultiMedia Sports scenario a little
more.  In thinking about the problem in general I understand the need
for data to be interchanged between portlets controlled by a consumer 
but haven't yet come up with good examples where events are needed.  The
current description of the scenario doesn't illuminate this.  Can
someone help me?  Because I think that if we can get away with only
needing to represent model/data we will end up with a simple API that is
clearer to use [in a manner that works efficiently].

2a.  I think we need to define/publish transparent state however I
don't 
think we need information related to scope/etc.  Is this there because
some folks want the consumer to manage this state for the producer 
outside the opaque mechanisms we have?  I think this would be an 
unneeded overcomplication. Why can't we get away with having a producer
merely publish a model of its public data and leave it to it to decide
how this maps to state/scopes?  I.e. A producer returns a representation
of this model on performBlockingInteraction. [As an optimization we can
have the consumer pass information to the performBlockingInteraction 
regarding what model elements it cares about/if any]. A consumer 
mediating state between itself/other portlets that receives a model back
from a performBlockingInteraction is able to push this state through to
other portlets via an UpdateModel call.  This call in turn returns
a 
model and the sequence continues until the consumer detects no more 
updates are needed or it detects a deadlock.

2b. Need for opaque portlet-defined events. I don't yet understand/see
a 
need for portlet defined events.  If we don't need them I also don't
understand why we need events at all.  Do we have enough if we just
focus on communicating state/data?

2c. Should we piggyback event delivery on existing operation signatures?
 Yes and no.  At this point [until generic events are justified] As
per 
above we need to extend the performBlockingInteraction call to carry 
information.  But we also need a new call to handle the dependent
updates [UpdateModel].

2d.  [New questions] My statements above assume that data changes
[events] never happen/get communicated out of band.  They only happen
in 
response to a user action.  Do we have requirements to allow the portlet
to trigger an out of band indication that state has changed?  [I hope
not].  How about an in-band one where you aren't the target of an
action 
or are asked to update your model?  [Only if we can live with allowing
a 
flag to be returned by render that indicates to consumers that can deal
with it that their model has changed, so that the consumer can choose,
if it wants to, to remember this and manage the state chanegd processing]

     -Mike-

richt2@us.ibm.com wrote:

>WSRP Coordination SC biweekly call has been modified by Rich Thompson
(richt2@us.ibm.com).
>
>Event description:
>USA Toll Free Number: 866-769-7222
>USA Toll Number: +1-203-566-0687
>PARTICIPANT PASSCODE: 377048
>
>Agenda:
>1. Considering the scenario assigned for driving SC work, the MultiMedia
Sports scenario, does the offered presentation capture the items we need
to consider?
>
>2. First set of questions to consider:
> - Need to define transparent publicized state?
> - Need for opaque portlet-defined events?
> - Should we piggy-back event delivery on existing operation signatures?
>
>
>Date:  Wednesday, 30 July 2003
>Time:  11:00am - 12:00pm Eastern Time
>
>This event is one in a list of recurring events.
>Other upcoming dates include:
>
>Wednesday, 09 July 2003, 12:00pm to 01:00pm Eastern Time
>Wednesday, 13 August 2003, 11:00am to 12:00pm Eastern Time
>Wednesday, 27 August 2003, 11:00am to 12:00pm Eastern Time
>Wednesday, 10 September 2003, 11:00am to 12:00pm Eastern Time
>
>
>View event details:
>http://www.oasis-open.org/apps/org/workgroup/wsrp/wsrp-coord/event.php?event_id=1670
>
>PLEASE NOTE:  If the above link does not work for you, your email
>application may be breaking the link into two pieces.  You may
be able to
>copy and paste the entire link address into the address field of your
web
>browser.
>
>
>
>You may leave a Technical Committee at any time by visiting http://www.oasis-open.org/apps/org/workgroup/wsrp-coord/members/leave_workgroup.php
>
>  
>



You may leave a Technical Committee at any time by visiting http://www.oasis-open.org/apps/org/workgroup/wsrp-coord/members/leave_workgroup.php