Why wouldn't the namespacePrefix be passed back to the producer as a
field [encodedNamespacePrefix?] of InteractionParams and/or
ResourceParams? Isn't all we are trying to do here is to formalize
encoding this prefix in interaction state so developers don't have to
figure out they need to use interaction state and we overcome the
limitations that such opaque encoding in interaction state may not work
in all situations nore that we have a corresponding facility in
ResourceParams?
-Mike-
Rich Thompson wrote:
I had also had concerns as I
contemplated
the tentative solution over the weekend.
If we add a portlet URL parameter to
carry the namespacePrefix used during markup generation, how does the
portlet
receive this back?
- one option would be to add
another
field to RuntimeContext carrying the previousNamespacePrefix, but this
raises both complexity and a question about why this would be distinct
from the current namespacePrefix.
- another option would be to
increase
the lifetime of the namespacePrefix to this particular usage of the
Portlet
by the Consumer. This would remove the need to store the prefix in the
URL as the portlet would also receive back the value used during the
encoding.
The downside is that the Consumer would either need to persist this
value
or have a dependable means of constructing its value.
In regards to mandating that
wsrp_rewrite_
not be used when encoding items that could flow back on interactions,
this
causes problems with using static script that references html form
fields
by name. One could argue that such script requires a rework already in
order to have the wsrp_rewrite_ inserted and that this rework could
instead
have the prefix supplied to the script on each invocation. In the
interest
of reducing complexity for the portlet (script) developer, I would
instead
argue for requiring that Consumer rewriting use the same value as is
supplied
to the Portlet with namespacePrefix. This also allows a broader use
case
of mixing dynamic (uses namespacePrefix) and static (uses
wsrp_rewrite_)
content.
Are there edge use cases that
wouldn't
be solved by this pair of changes?
1. Increase the lifetime of the
namespacePrefix
to the particular usage of the Portlet by the Consumer.
2. Require that Consumer namespace
rewriting
use the same value as what is supplied to the Producer in
namespacePrefix
I also think there is value to
making
namespacePrefix and portletInstanceKey required fields (Producer can
depend
on them) and am unaware of any use cases where the Consumer would have
a difficult time supplying these values. In general this would be a
case
of v2 raising the bar of what constitutes a good Consumer
implementation.
Rich
Here goes the proposal I had, just wanted to
clarify
it so people can make
up their mind.
Rationales:
- do not require form params namespacing
- do not prevent portlets from doing it (but don't add an extra penalty
on
those who don't want to do it)
- do not change portlet programming models, allow them to use the means
they have today
- do not require a constant prefix accross requests (although we coud
talk
about it, portletInstanceKey is already someting similar)
- have a symmetric model, i.e. portlet gets back what it encodes
- try to avoid unnecessary rewriting steps
The whole proposal simply bases on two facts:
- why not force consumers to always provide a namespace prefix? today
it
is
optional
- using wsrp_rewrite_ for form param names is problematic and should
not
be
used (we discussed it already)
1. Consumer always provide namespacePrefix
This way Producer can rely on it and always return the prefix to the
portlets if the request it, e.g. they call getNamepace() in the JSR168
case.
I'm pretty sure other APIs will have similar means.
There is no real penalty on the consumer, they can easily compute one
per
page AND they already need some means for themselves anyway.
So here portltes can encode their form field names using that namespace.
The only thing they need is to remember the prefix they used.
This can easily be stored in the portltet's interactionState.
This is completly transparent to the consumer AND producer and doesn't
require any additional rewriting, parsing, stripping etc.
2. do not use wsrp_rewrite_ in form parameter names
a) there is no need for it if we have 1.
b) as said it is very problematic, and was not the initial intent of
wsrp_rewrite_. The intent was to namespace *one time* the markup
without
a
flow back. It seems the stretch we want to do here is to large.
c) in fact portlets using getNamepace() and 1. will never see a
wsrp_rewrite_ anymore, so from a dynamic UI generation point of view
there
is no need to use wsrp_rewrite_
d) portlets with static content can still use wsrp_rewrite_ as they use
for
1.0 we don't break anything
3. the resource discussion we had last thursday
The proposal on the table doesn't solve the problem anyway as I
described
below.
Let's think through it:
- in 1.0 it never could work and indeed we discouraged people from
namespacing form field names.
- if a resource (e.g. servlet) uses wsrp_rewrite_ we really can
consider
it
as being "WSRP aware", i.e. we can't claim that in this case
the resource
is completly WSRP agnostic.
- if the above is true why not pass the "WSRP context" to the
resource via
the resource URL. In this case why not pass the namespacePrefix to the
resource?
This is how it could work:
1. portlet gets called and obtains the namespacePrefix
2. portlet encodes a resource URL and as a parameter passes the
namespecePrefix to be used (indeed it needed to be a subSpace of the
initial namespace in the case the portlet AND the resouce generate
forms
-
use case?)
3. resource received ns prefix can use it, can also encode the prefix
used
by various means.
Mit freundlichen Gruessen / best regards,
Richard Jacob
______________________________________________________
IBM Lab Boeblingen, Germany
Dept.8288, WebSphere Portal Server Development
WSRP Team Lead & Technical Lead
WSRP Standardization
Phone: ++49 7031 16-3469 - Fax: ++49 7031 16-4888
Email: mailto:richard.jacob@de.ibm.com
Richard
Jacob/Germany/IBM
@IBMDE
To
wsrp-interfaces@lists.oasis-open.or
09/06/2005 11:17
g
AM
cc
Subject
Re: [wsrp-interfaces]
interactionFieldPrefix
and
namespacing
form params
Hi all,
I just decided to stress you further on the topic what we discussed last
week :-)
I started to think through what we proposed and decided to give it a
shot.
So here is what I understood as the general idea, please correct me if
I'm
wrong:
- introduce a new WSRP url parameter containing the namespace which was
used by the producer container and passed to the portlets, e.g. if they
call getNamespace()
- we assumed implicitly that the value of the new param is flowing back
(in
InteractionParams and MarkupParams?)
What does it bring/help or was the intent of it?
- do not require namespacePrefix to be mandatory
- allow usage of wsrp_rewrite_ as the namespace prefix, here we
explicitly
refered to resources using rewriting
- do not require the namespacePrefix to be constant across requests
- anything I missed?
I've been playing a little bit with it, here are some thoughts
1. portlets do not use wrsp_rewrite_ for form params namespacing
- it seems we're introducing some kind of namespace persistance across
requests through the backdoor.
Once the producer container decides to pick up a namespacePrefix from
the
runtime context it is "persisted" in the URLs.
The Consumer is forced to send it back -> Producer uses it again and
stores
it in the URLs.
While this seems an interesting means at the first glance, it introduces
some problems.
- How can consumers guarantee that they do not assign a new
namespacePrefix
on a page which clashes with the existing ones?
- How does this new value relate with the existing namespacePrefix,
seems
quite confusing here? Is the namespacePrefix assigned the new value from
the URL?
2. portlets using wsrp_rewrite_ in form names as the ns prefix
We saw that this is quite problematic. The intent of the proposal using
the
URL param was that the URL param itself gets rewritten by the consumer
to
the "real" namespace as well as the form fields in the markup.
Thus
portlets/the producer can obtain the consumer assigned namespace to
identify the form params.
- Do we expect that Producers in this case pick up the assigned namepace
and rewrite it back again to wsrp_rewrite_ so that rewriting is always
used? -> additional producer rewriting
- If not we have just one wsrp_rewrite_ usage and then fall back to a
given
namespace prefix with the implication I described in 1. ?
- I think we change the rewriting algorithms here: What do consumers do
if
the wsrp rewrite token appears as a parameter value in the rewrite URL?
Aren't parameter values opaque here? Do your rewriting implementation
touch
parameter values in a rewrite URL?
3. Resources using URL rewriting
I think the proposal doesn't really solve the problem. The main question
here is:
- How does the namespace prefix assigned by the consumer flow back to
the
reasource?
- If it needed to flow back to the reasource it needs needs to be part
of
the resource URL not the rewrite URL. But as of today, resource URLs are
opaque to the Consumer and it shouldn't touch them.
So when thinking through it, things really seem to become clumsy here,
too
with not adding to much value here.
There are some implication in the proposal, which haven't been thought
through, yet.
Mit freundlichen Gruessen / best regards,
Richard Jacob
______________________________________________________
IBM Lab Boeblingen, Germany
Dept.8288, WebSphere Portal Server Development
WSRP Team Lead & Technical Lead
WSRP Standardization
Phone: ++49 7031 16-3469 - Fax: ++49 7031 16-4888
Email: mailto:richard.jacob@de.ibm.com
|