Alex,

I suppose the problem is going to be that the components are configurable to
the level which can affect the input/ouput,  and therefore it is the final
implementation that needs to generate the WSDL.

A component like the Groovy Component is going to be difficult to provide
WSDL for - it is the responsiblity of the implementor.

I can see the WSDL 1.1 would work for stricter interfaces (such as those of
the JSR181 component) but what about more flexible content?

P

On 7/20/06, Alex Boisvert <[EMAIL PROTECTED]> wrote:


My suggestion would be to go towards WSDL 1.1.  It's a widely accepted
spec with well-defined rules of interoperability if you take into
account the WS-I BasicProfile 1.1.  And we have a good mapping defined
in the JBI spec for passing around WSDL 1.1 normalized messages.

I'm not a WS fanatic but I don't see any other approach that would
better formalize the service invocation contract (esp. the normalized
message format) between components.   WSDL also provides the ability to
map header properties into the (JBI) message content as required by the
use-case Ramon described below.

It's a little more work for each component to define a WSDL 1.1 mapping
but in the end you get much more value out of the whole system because
all components can work together out-of-the-box without worrying about
transformations, ad-hoc mappings, adapters, etc.

And after all, it's the big idea behind the separation of binding
components and service engines.

alex


Philip Dodds wrote:
> Its certainly a good point, so far most of the components have been
> pretty
> self-centred and the specification is purposefully vague.
>
> Perhaps it is time to start looking at message sterotypes of some kind
to
> allow people using servicemix a little better understand of the
> input/output
> message types.  This might at least provide some guidance as to what
fits
> together out of the box and what might need transformation?
>
> I wonder if we should look at extensing a wiki page or two on some
> ServiceMix standards?  Could be as little as a list of message types
> and we
> could provide either an index of the components by type?
>
> P
>
> On 7/19/06, Ramon Buckland <[EMAIL PROTECTED]> wrote:
>>
>> Interesting topic.
>>
>> In our recent development work
>> we chose BPEL as part of an implementation
>> but BPEL cannot access JBI properties, which when we spec'd our
>> work, were going to be used as META data for the messages.
>>
>> As it turned out, we came up with our own message format
>>
>> <rootNode>
>>     <properties>
>>         <propertyName>value</propertyName>
>>       ...
>>     </properties>
>>     <payload>
>>         .. actual XML ..., data, command request etc
>>     </payload>
>> </rootNode>
>>
>> and we marshal between our bespoke and JBI NormalizedMessage
>> when we need. (ie, shift properties to and from the message content
>> to NM properties when needed).
>>
>> It posed some tricky decisions. Going forward for the use
>> we have, this custom normalisation is good, (meaning we have one WSDL
>> for
>> BPEL messaging
>> etc .. ).
>>
>> For Eg: when -http provides the response .. regardless of it's
>> normalisation
>> we run it through a custom "normaliser" and then in BPEL would get
>>
>> <rootNode>
>>     <properties>
>>       ...
>>     </properties>
>>     <payload>
>>         .. actual XML from servicemix-http ..
>>     </payload>
>> </rootNode>
>>
>> our two cents.
>>
>>
>>
>> On Wed, 19 Jul 2006 07:50:15 -0700, Alex Boisvert wrote
>> > To tell you the truth, I was secretly hoping to spur a debate around
>> > message normalization.  :) The way I understand it, if I start
>> > changing the message format put on the bus, it will most likely
>> > break other components that expect the older format.
>> >
>> > I'd be curious to hear what other think about this.  Various
>> > components seem to use different normalization rules (or no
>> > normalization at all) which will inevitably lead to interoperability
>> > problems down the road.   Time to set higher standards?
>> >
>> > alex
>> >
>> > Philip Dodds wrote:
>> > > Alex,
>> > >
>> > > I don't believe anyone is,  and we would more than welcome a
>> patch :)
>> > > just
>> > > create a JIRA and attach it
>> > >
>> > > Thanks
>> > >
>> > > P
>> > >
>> > > On 7/18/06, Alex Boisvert <[EMAIL PROTECTED]> wrote:
>> > >>
>> > >>
>> > >> Hi,
>> > >>
>> > >> I've noticed that servicemix-http simply places the child
>> element of
>> the
>> > >> SOAP:Body as the content of JBI normalized message.  This doesn't
>> seem
>> > >> to go with the spirit of normalization... I would have expected a
>> WSDL
>> > >> 1.1-wrapper element with message parts if I deployed a WSDL 1.1.
>> > >> document.
>> > >>
>> > >> Is anybody working on this yet?  If not, I could volunteer for a
>> patch.
>> > >>
>> > >> cheers,
>> > >> alex
>> > >>
>> > >>
>> > >>
>> > >>
>> > >>
>> > >
>>
>>
>>
>> --
>> Open WebMail Project (http://openwebmail.org)
>>
>>
>


Reply via email to