I do see the usefulness of a limited consumer
to portlet value sharing, with the consumer supplying values for “slots”
requested by a portlet with no push back by the portlet. Anything that updates values
to be held by the consumer (e.g. the shared timezone) to me implied that the
changes be propogated to the consumer’s store and to other portlets to be
useful. Maybe that is all we can support in 2.0 (with some error feedback
channel as I previously suggested).
Regards,
Andre
From: Michael Freedman
[mailto:michael.freedman@oracle.com]
Sent: 27 July 2005 13:13
To:
wsrp-interfaces@lists.oasis-open.org
Cc: Michael Freedman
Subject: Re: [wsrp-interfaces] Fw:
[wsrp] Additional use cases for issue #44 (set new public params)
Yes, I am in the U.K.
this week enjoying the sunny weather :-)
The base PP use case is one where the consumer wants to [in part] drive the
content the portlet produces. The simple example I used way back when was
a portlet that was locale [zip code] specific such as the weather portlet.
By publishing the PP for zip code the consumer could drive the content of
this portlet -- for example by using information from the current user's user
profile, or using a predefined [page] value, or even a value it
publishes/recognizes from URLs submitted by its clients. In each of these
cases the value is commonly request dependnent. In this base scenario the
PPs are local to a portlet [each portlet's are independent]. Coordination
is established/managed by the consumer when it decides to use the same values
for PPs in differing portlets. I.e. this mechanism allows a Consumer to
coordinate with a producer. What it doesn't do is allow one
producer/portlet to coordinate with another. We have events for this.
Currently, we are discussing expanding PPs so they can also be used as an
optimization mechanism for portlet to portlet coordination. I am okay
with us doing that if we don't expand/change the base semantics of PP [i.e. its
public navigational state] and we prefer this mechanism over a type of
transient property mechanism we might add in the future. Personally, I
currently think transient properties will be the better/best solution, once we
define them and that we should wait vs. using PPs so as not to confuse folks
later by giving them too many options for doing roughly the same thing. I
don't see that our lack of support for this optimization in 2.0 would be bad --
the function is doable via events just a little less efficiently.
My comments on PerformInteraction were not really in relation to PPs. I
am trying to separate these some. My PerformInteraction comment suggests
rather then overload getMarkup and misshandling the definition we already seem
to have defined a mechanism that fits the purpose proposed. If so then we
should use it. When the follow-on question arises of how do I encode/do
cross portlet coordination with a performInteraction URL -- my answer looks
back to the independent decision we made above. If we decide that
transient properties is the right way to do this, then in WSRP 2.0 the answer
is you can't -- you either wait for us to add transient properties or you use a
BlockingInteraction/event. If our answer to the above is we are willing to
live with the limited set of semantics you get when you define Public
Navigational state where consumers are required to coordinate providing a
consistent value when like names are used -- then you have the support we will
add for setting/encoding PPs.
-Mike-
Andre Kramer wrote:
Hi
Mike (you must be on UK
time),
Could you describe a use case were the PP
is just shared between the consumer and one portlet? The only ones I can think
of are really scenarios for a complex consumer (also know as rich or smart
client) that should be doing Web Services calls that are application specific
rather than piggy backing on a generic protocol like WSRP.
My confusions about the separate issues
may be caused by not seeing uses that are not communicated between portlets
(what I would call shared but non-public) or not maintained between requests
(what I would call transient). PPs promised to make coordinated values (such as
current time zone communicated to multiple portlets) easier for portlet programmers
than full blown event handling but seem to require that they be public and
non-transient.
Thinking ahead, relaxing the blocking
semantics of actions would help but would this not take the event handling
communication round out of such interactions hence forcing (or encouraging
depending on viewpoint) use of PPs? I think I would rather encourage use of
events and discourage PPs.
Regards,
Andre
How did we take two separate discussions/issues, fuse them into one and
make the resulting discussion even more cloudy?
Anyway, on PP let's get back to basics.
Didn't we decide a few weeks back that we were not going to design/describe
transient properties in this realease?
Didn't we decide that PP were a separate concept from transient properties?
And that the distinction between the two is that PP is public
navigational state with no defined consumer scope while transient properties
are consumer managed state with defined scope?
And don't we define public navigational state semantics as equivalent to
regular navigational state semantics except that the value is not opaque to the
consumer?
So isn't the only issue here whether the use cases Stephan and Richard have
given us are reasonable/preferrable to solve via PP [as compared to using
either events or the to be designed transient properties]?
If so, here are my comments...I think we need to compare constrast supporting
these use cases with PP vs transient properties. PP doesn't introduce any
notion of shared state -- each portlets navigational state is its own.
Transient properties will likely support a notion of shared state.
Yes, we can add some conventions to PP to require the consumer to associate
PPs with the same name meaning they should receive the same values -- and this
is what we have suggested we use to support the use cases. If we want
this semantics then, yes we need to offer equivalent mechanisms to the portlet
to set public navigational state as we do regular navigational state. I
think Rich's proposal does this. If we end up thinking that once we have
transient properties it will make more sense to support these use case with
them [since the scope of sharing will likely be defined/influenced by the
portlets] then I would suggest it might be better if we didn't add this
behavior to PPs and merely leave them restricted to the defining portlet.
If we choose this later approach then we don't need any way for the portlet
to set PPs as setting regular navigational state should suffice. [And
remember, the use cases are supportable in 2.0 via events while we wait for a
more efficient mechanism in 3.0].
As for supporting method GET for render URLs, as I said last week, it seems
what is being proposed here is exactly what performInteraction was all about.
Why isn't this merely a proposal that we add performInteraction back into
the specification? Rich, can you dust off the old performInteraction
description and send it out to folks so we can see if we already have a
completely designed and if I recall agreed upon set of semantics?
-Mike-
Rich Thompson wrote:
I'm sorry, but I find the whole SOAP vs REST debate irrelevant to this
discussion.
There
are some key questions though:
1.
Does the semantics of getMarkup match that of HTTP GET so that Consumers can
use GET for render links?
If the
answer to #1 is yes (I think it is), then:
2. If
PPs are added to URLs, can the Consumer support them for the scope of the user
interaction across all portlets on the page without violating HTTP GET
semantics?
3. If
PPs are added to URLs, can the Consumer support them for a scope encompassing
multiple user interactions across all portlets interacted with and not violate
HTTP GET semantics?
I
think the answer to all three questions is yes, though #3 requires the Consumer
to support PPs as part of its own navigation state. Common web usage would also
allow storing the PPs in a user session, though technically this would violate
GET semantics as it is an update of server-side state.
Rich
But surely the (if any) desire is to enable GET
semantics between the browser and the consumer Portal for optimized
“render” interactions? All WSRP operations today use SOAP (and
therefore POSTs) and are explicitly not REST. In any case, I would argue for a
simple developer model that supports the use cases. These, for me, naturally
require a statefull consumer to serialize updating changes to shared public
parameters or could possibly be poorly supported by a set-for-one-interaction
/one-shot imposed limitation. I don’t particularly wish to go down the
“Web GET/POST semantics” rat hole for WSRP just for such a
limitation.
Regards,
Andre
From: Rich Thompson [mailto:richt2@us.ibm.com]
Sent: 26 July 2005 14:15
To: wsrp-interfaces@lists.oasis-open.org
Subject: RE: [wsrp-interfaces] Fw: [wsrp] Additional use cases for
issue #44 (set new public params)
I can definitely see useful coordination happening without the reversion you
describe if the Consumer chooses to treat PPs as part of the Consumer's
navigational state. This keeps it from impacting state managed server-side
(i.e. is compatible with GET semantics) while enabling ongoing, state-oriented
coordination of the portlets being rendered for the user (i.e. once the zipcode
PP is set, it stays at that value until it is reset by some other
user/Consumer/portlet action).
Rich
Could Stefan & Richard please remind us as to when it is useful to have a
render URL affect other portlets on a page (just) for the duration of one page
fetch, when a subsequent refresh (browser button) is likely to revert to the
previous consumer state (as nothing permanent is being implied spanning
consumer/portlets)?
I suspect any useful coordination state change seems more related to a consumer
effectively tracking and forwarding shared values and so the Web request to the
consumer should itself be a POST to the consumer/Portal? Given PPs on render
URLs that don’t last v.s. PPs in action replies that do (SHOULD/MUST)
aren’t developers going to choose actions anyway, e.g. updating a
charting page’s units from imperial to metric?
Regards,
Andre
From: Rich Thompson [mailto:richt2@us.ibm.com]
Sent: 25 July 2005 15:02
To: wsrp-interfaces@lists.oasis-open.org
Subject: Re: [wsrp-interfaces] Fw: [wsrp] Additional use cases for
issue #44 (set new public params)
I think it might be time to pull this discussion back out of the abstract space
and relook at the use cases we have accepted as compelling.
Stefan & Richard have posted some use cases that clearly would benefit from
PPs being settable on URLs and need a PP scope of at least the user interaction
... including all portlets accessed during the processing of the user
interaction. Could you post use cases requiring broader scopes? I presume that
if we find them compelling, my answer to Mike's previous inquiry about
requiring Consumers to support PPs may need to change.
Rich
Stefan,
This is question of scoping.
As we have been debating, there are two options to let portlets update
some shared state (public params). The first option is to let portlets
encode such change in URLs, and the second is to let producers return
new public params within SOAP.
Apart from the encoding rules, these options influence scoping of this
shared state that consumers can implement. If you go with the first
option, consumers will have to limit this shared state to a user's
session or just a few portlets on a given page. IMO, this is restrictive.
However, if you go with the second option, consumer has more choice.
Consumers can implement wider scoping of this shared state.
Subbu
Stefan Hepper wrote:
> I think this is mixing sematics. Anything that requires a pbia is ment
> to be a resource state change. For resource changes you need a blocking
> semantic. Navigational state and public params in my mind define the
> view state of the portlet and changing this state does not require a
> pbia nor a non-blocking action. The new parameters or state can just be
> provided to the portlet.
>
> This also relates to the W3C standard about GET and POST:
>
> " * Use GET if:
> o The interaction is more like a
question (i.e., it is a safe
> operation such as a query, read operation, or lookup).
> * Use POST if:
> o The interaction is more like an
order, or
> o The interaction changes the
state of the resource in a way
> that the user would perceive (e.g., a subscription to a service), or
> o The user be held accountable
for the results of the
> interaction."
>
> I don't see why it is necessary to pay the penalty of an action for
> changing PPs.
>
> Stefan
>
>
>
> Subbu Allamaraju wrote:
> > Rich Thompson wrote:
> >
> >>
> >> My comment was that requiring action processing is always
possible,
> >> but why would we require it in this case and not for
> >> mode/windowState/navState changes?
> >
> >
> > I agree, but the difference is the encoding rules.
> >
> > Subbu
> >
> >>
> >> Rich
> >>
> >>
> >> *Subbu Allamaraju <subbu@bea.com>*
> >>
> >> 07/22/05 09:48 AM
> >>
> >>
> >> To
> >> wsrp-interfaces@lists.oasis-open.org
> >> cc
> >>
> >> Subject
> >> Re: [wsrp-interfaces] Fw: [wsrp] Additional
use cases for issue
> >> #44 (set new public params)
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >> Thinking aloud, can't we map render URL use cases to use
pbia? The
> >> reason for changing the value of a paramter could be the
result of a
> >> user interaction, which can then compute new values and
return with
> pbia
> >> response.
> >>
> >> Subbu
> >>
> >>
> >> Rich Thompson wrote:
> >> >
> >> > It is possible to drive every style of update
through action
> >> processing,
> >> > but I think that short changes the advantages of
using render URLs
> >> where
> >> > they are appropriate.
> >> >
> >> > New navState can be specified on each URL
(including render
> URLs), why
> >> > should not the Portlet be able to specify the
shared version (PPs)?
> >> >
> >> > Rich
> >> >
> >> >
> >> > *"Andre Kramer" <andre.kramer@eu.citrix.com>*
> >> >
> >> > 07/22/05 08:02 AM
> >> >
> >> >
> To
> >> >
Rich Thompson/Watson/IBM@IBMUS,
> >> <wsrp-interfaces@lists.oasis-open.org>
> >> > cc
> >> >
> Subject
> >> >
RE: [wsrp-interfaces] Fw: [wsrp] Additional use
> >> cases for issue #44
> >> > (set new public params)
> >> >
> >> >
> >> >
>
> >> >
> >> >
> >> >
> >> >
> >> > After reading your previous email, I was driving
towards a similar
> >> > recommendation for Portlets to
“re-fresh” public parameters by
> >> return on
> >> > each WSRP interaction so that a consumer is more
likely to forward
> >> > values that have been recently actively returned
and I agree this
> >> means
> >> > Portlets have to maintain/store public param
values e.g. in
> >> navigational
> >> > state. But, with such advice, I see no reason to
have public
> >> parameters
> >> > to visibly appear on URLs (rewrite expressions
or templates) at
> >> all. Why
> >> > introduce a new encoding and (in markup)
transport mechanism when it
> >> > seems a preferable Web Service (in SOAP
response) encoding and
> >> transport
> >> > is already available and recommended?
> >> > > Regards,
> >> > Andre
> >> > >
> >> >
> >>
> ------------------------------------------------------------------------
> >> >
> >> > *From:* Rich Thompson [mailto:richt2@us.ibm.com] *
> >> > Sent:* 22 July 2005 12:24*
> >> > To:* wsrp-interfaces@lists.oasis-open.org*
> >> > Subject:* Re: [wsrp-interfaces] Fw: [wsrp]
Additional use cases for
> >> > issue #44 (set new public params)
> >> > >
> >> > The Interfaces SC discussion raised the question
about the scope of
> >> PPs
> >> > set in this manner. Namely; does the Portlet
need to keep
> >> specifying the
> >> > PPs on all subsequent requests or can it depend
on the Consumer to
> >> > resupply those values in the future (i.e. some
scope, like
> >> user-session,
> >> > becomes mandated).
> >> >
> >> > I think we should be careful to leave scoping
questions up to the
> >> > Consumer, but if PPs are to function as a
coordination model,
> then we
> >> > also have to discourage Portlets from setting
all PPs on every
> URL. A
> >> > model where Portlets are encouraged to store
current PP values
> >> > (preferably in navState) for use as default
values should the
> Consumer
> >> > not supply a value on a subsequent invocation
and only do PP setting
> >> > when a value needs to change accomplishes this.
It removes
> setting the
> >> > value each time such that coordination happens
in a more reasonable
> >> > manner while also providing for reasonable user
experience, even
> when
> >> > the Consumer does not support and PP scope
beyond the user
> >> interaction.
> >> >
> >> > Note: We could consider a model where Portlets
MUST encode all
> PPs on
> >> > all URLs such that this supplies the scope,
rather than the Consumer
> >> > determining the scope, but that this fails as
the Portlet might not
> >> have
> >> > access to all PPs (security and privacy
reasons).
> >> >
> >> > Rich
> >> >
> >> > *Rich Thompson/Watson/IBM@IBMUS*
> >> >
> >> > 07/07/05 11:54 AM
> >> >
> >> >
> To
> >> >
wsrp-interfaces@lists.oasis-open.org
> >> > cc
> >> >
> Subject
> >> >
[wsrp-interfaces] Fw: [wsrp] Additional use cases
> >> for issue #44 (set
> >> > new public params)
> >> >
> >> >
> >> > >
> >> >
> >> >
>
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > Reposting, as requested ...
> >> > ----- Forwarded by Rich Thompson/Watson/IBM on
07/07/05 11:52 AM
> -----
> >> >
> >> >
*Rich Thompson/Watson/IBM@IBMUS*
> >> >
> >> > 06/15/05 09:24 AM
> >> >
> >> >
> To:
WSRP TC
> >> <wsrp@lists.oasis-open.org>
> >> > cc:
> Subject: Re: [wsrp]
Additional use
> >> cases for issue #44 (set
> >> > new public params)
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > I had taken a to-do from the Interfaces SC to
develop (in
> conjunction
> >> > with Stefan, Richard and Mike) a proposal for
this portion of Issue
> >> #44.
> >> > Here is that proposal: *_
> >> >
> >> > Philosophy:_* Since Public Parameters (PP) are
another aspect of
> >> > Consumer managed state that is exposed to
Portlets (the other two
> are
> >> > modes and windowState), Portlets should be able
to request
> changes to
> >> > PPs in much the same way as they do modes and
windowStates (on URLs
> >> and
> >> > in responses from pbia and handleEvents). *_
> >> >
> >> > On URLs: (2 issues)_*
> >> > 1. Need to encode 3 pieces of information
(request to set a PP, the
> >> > PPname and the PPvalue) into 2 places
(name=value) when using a
> >> > querystring and deal with the reduced set of
characters allowed
> in the
> >> > path portion ('=' is not allowed). Templates
also introduce an issue
> >> > with how to encode multiple PPs ... preferably
with the Template
> only
> >> > having a single placeholder.
> >> > -Proposal:
> >> > 1. All public parameters specified on a URL are
concatenated in the
> >> form
> >> > of
"PPname1=PPvalue1&PPname2=PPvalue2".
> >> > 2. The resulting string is URL encoded (changes
'=' into %3D and '&'
> >> > into %26) to make it valid in both the
querystring and path
> >> portions of
> >> > a URL.
> >> > 3. This URL encoded string becomes the value for
the
> >> > wsrp-publicParameter URL parameter, regardless
of whether template
> >> > processing or Consumer URL rewriting is in use.
> >> >
> >> > 2. How to encode complex PPvalues? I suggest
serializing the
> >> PPvalue to
> >> > XML and URL encoding this XML. The Consumer would
then receive
> the PP,
> >> > recognize it is of a complex type (based on
PPname) and decode the
> >> > PPvalue to get the XML. *_
> >> >
> >> > On Operation responses:_*
> >> > 1. Return PP requests via a field much as
newMode and newWindowState
> >> > request updates to those aspects of Consumer
managed state. Suggest
> >> this
> >> > field be of type QNamedStringArray, though it
could be an array of
> >> > NamedString if we do not want to introduce a
usage of a type from
> the
> >> > 'extra' namespace.
> >> >
> >> > Rich
> >> >
> >> > *Michael Freedman <michael.freedman@oracle.com>*
> >> >
> >> > 05/24/05 07:55 PM
> >> >
> >> >
>
> >> >
> >> > To
> >> >
WSRP TC <wsrp@lists.oasis-open.org>
> >> > cc
> >> >
> Subject
> >> >
Re: [wsrp] Additional use cases for issue #44 (set
> >> new public params)
> >> >
> >> >
> >> >
> >> >
> >> >
>
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > I too am interested in hearing people's current
opinions. Our early
> >> > coordination discussions considered this
carefully. It was
> decided at
> >> > that time to not support two coordination models
that delivered
> >> > equivalent function. Though I pushed
strongly for such a parameter
> >> > style mechanism, the subcomittee preferred
Events because it not
> only
> >> > allowed state to be transferred but also
actions. The current
> public
> >> > parameter model was introduced by me later and
was crafted
> >> specifically
> >> > to not step on the the toes of Events.
>
> >> > To extend this conversation I would like to
pose:
> >> > 1. We consider
whether publicParameters should be QNames so
> >> they
> >> > can be shared/reused across producers.
Note: this is likely needed
> >> > whether we stick with events or add the ability
to encode a public
> >> > parameter directly in the URL.
> >> > 2. We consider adding
a new "defined event" called
> >> > publicParameterValueChanged. The payload
of this event would be the
> >> > QName of the parameter and an Object any holding
the parmeter value
> >> > [though it might be nice to support our
Any/String optional pair
> style
> >> > here].
> >> > 3. We consider
defining the meaning of publicParameter whose
> >> > capability contains the value "required"
as meaning that normal
> >> usage of
> >> > this portlet requires the consumer to provide
this value. The
> portlet
> >> > would still have to deal with situations in
which this wasn't
> provided
> >> > but likely an end-user would consider this usage
> crufty/abnormal. For
> >> > example if a portlet required a CustID parameter
and didn't receive
> >> one
> >> > it could display a view that asks for a custID.
The key here by
> >> saying
> >> > "required" [we can choose a different
capability name] the consumer
> >> can
> >> > distinguish between those public parameters that
have a secondary
> >> impact
> >> > on the portlet [optional] and those that have a
primary or important
> >> > impact [required].
> >> > -Mike-
> >> >
> >> > Stefan Hepper wrote:
> >> > Hi,
> >> > I've some use cases that may be a good match for
the ability to set
> >> new
> >> > public parameters by the producer and to encode
public parameters in
> >> > URls by the producer.
> >> >
> >> > 1. displaying content based on a specific
product id:
> >> > - Portlet A allows to select a product from a
list
> >> > - User clicks on a specific product
> >> > - Portlet B renders details of this product
> >> > - Portlet C renders currently available number
of items on stock for
> >> > this product
> >> > - user wants to bookmark this result in order to
come back to
> product
> >> > tomorrow
> >> >
> >> > implementation with events:
> >> > Portlet A encodes the customer selection URLs as
POST action links
> >> > Portlet A receives a performBlockingInteraction
call
> >> > Portlet A returns event productID=10
> >> > Portlet B receive a blocking handleEvents call
for productID=10 and
> >> > returns new navigational state
> >> > Portlet C receive a blocking handleEvents call
for productID=10 and
> >> > returns new navigational state
> >> >
> >> > implementation with public params:
> >> > Portlet A encodes the product selection URLs as
GET render links
> with
> >> > the productID as public param
> >> > Portlet A receives a render call with the public
param productID=10
> >> > Portlet B receives a render call with the public
param productID=10
> >> > Portlet C receives a render call with the public
param productID=10
> >> >
> >> > which would be much more efficient and also
consistent with the W3C
> >> > architecture as links that do only change view
state should be
> encoded
> >> > as GET links.
> >> >
> >> >
> >> > 2. display content based on a specific customer
id
> >> >
> >> > 3. display content based on the selected state
of a map
> >> > portlet A displays a map of USA
> >> > portlet B displays information on the selected
state (# people
> >> > registered, capital, ...)
> >> > portlet C displays all IBM labs in that state
> >> >
> >> > and many more.
> >> >
> >> >
> >> > This would allow to have some gobal navigational
state that can
> be set
> >> > via URLs by portlets. Also portlets may want to
set new public
> >> params as
> >> > a result of an blocking interaction or a handle
event call.
> >> >
> >> >
> >> > What do you think?
> >> >
> >> >
> >> > Stefan
> >> >
> >> >
> >> >
> ---------------------------------------------------------------------
> >> > To unsubscribe from this mail list, you must
leave the OASIS TC that
> >> > generates this mail. You may a link to
this group and all your
> TCs in
> >> > OASIS
> >> > at: _
> >> >
> >>
> __https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php_
> >> >
> >> >
> >>
> --------------------------------------------------------------------- To
> >> > unsubscribe from this mail list, you must leave
the OASIS TC that
> >> > generates this mail. You may a link to this
group and all your
> TCs in
> >> > OASIS at:
> >> >
> https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
> >> >
> >>
> >>
> >
> >
>
>