Gianfranco,

have a look at the http-upload tutorial. You will find that there is a
custom marshaler used.

Regards,
Lars



Gianfranco Boccalon schrieb:
> Hi,
> I'm trying to do the same thing but I dont find documentation on how
> to use a custom marshaller: in fact I dont find documentation on how
> using HttpConsumerEndpoint, because the configuration explained in the
> samples creates an HttpComponent that uses a HttpEndpoint, that is not
> configurable because it uses always the JBIMarshaller.
>
> Which is the configuration for using HttpConsumerEndpoint and then use
> a custom marshaller ?
>
> Thanks in advance
> Gianfranco Boccalon
>
>
>
>
> adam.strickland ha scritto:
>> I'm implementing an HttpConsumerEndpoint using a non-SOAP protocol. 
>> What I
>> did was create a custom HttpConsumerMarshaler to handle the inbound
>> request
>> (if you're interested, it a MIME Multipart-Related request;
>> significant in
>> that it is absolutely NOT XML compliant, which causes the various XML
>> parsing routines to barf) and configured the endpoint to use it.  In
>> this
>> use case, the external entity (i.e. an HTTP client) is expecting a
>> standard
>> HTTP Request-Response cycle, which corresponds to an InOut MEP;
>> therefore I
>> configured the endpoint to use InOut as the default MEP.  What I
>> found was
>> that the client was receiving an empty response, which did not
>> satisfy the
>> business requirements (the client expects an ACK or NAK type of message;
>> also MIME Multipart-Related, by the way).
>>
>> What I ended up doing was simply commenting/removing a line of code
>> in the
>> HttpConsumerEndpoint, recompiling and voila!  My custom version of
>> ServiceMix handled the request like a champ.  What I did was simply
>> remove
>> the 'return' statement (line 244), allowing the method to run to
>> completion
>> where it actually performs the send() of the outbound message.
>>
>> So here's the actual question: am I missing something here?  I didn't
>> see
>> another way around this, but I could just be blind to one of the many
>> implications or just have overlooked another component, etc. that may
>> solve
>> the problem better.
>>   
>
>

Reply via email to