You are viewing a plain text version of this content. The canonical link for it is here.
Posted to soap-dev@ws.apache.org by Paul Fremantle <pz...@hursley.ibm.com> on 2000/09/06 22:12:26 UTC
Accessing instances in SOAP - ideas?
Jackson, SOAP gurus!
I've been thinking about this (how to specify instances) for EJB instances.
I've come up with two approaches. One is to call the EJB home to get the a
URI for the instance. This has the benefit of not requiring special headers.
The other is to put an instance identifier in a header. I'm interested in
comments on both approaches, and any other approaches?
I've prototyped both, and the Java Apache client code in each case is ok. I
haven't really thought through lightweight clients and other clients.
Regards,
Paul
----- Original Message -----
From: "Jackson George" <Ja...@necx.com>
To: <so...@xml.apache.org>
Sent: Wednesday, September 06, 2000 4:31 PM
Subject: Jackson : How to Write the Server Side SOAP servlet to Communicate
to the the STOCK Quote
>
>
> Hi
> We are trying to implement SOAP with Java, But for 'stockquoteClient'
,
> how to write
> a Java Servlet that could communicate to the EJB and Call a Method, What
I
> have to do for this
> Server side Servlet. I know that Using RMI-from this Servlet I could
> communicate to the EJB.
> But Anybody COULD PLEASE send me a Step by step Instructions How to set
up
> the Client
> and How set-up the SOAP server , Please send me the Code too.
>
> http://www.xmethods.com/detail.html?id=2\
>
> I am a beginner in SOAP, Just a Beginner Hope you could help me
> Appreciate , Thanks in advance
> Jackson George
> jgeorge@necx.com
>
> -----------------------------------
>
> -----Original Message-----
> From: Jacek Kopecky [mailto:jacek@idoox.com]
> Sent: Tuesday, September 05, 2000 6:39 PM
> To: 'soap-dev@xml.apache.org'
> Subject: RE: IRC log
>
>
> Just a few quick remarks before I go to sleep. 8-)
>
> I first read the SOAP spec. early this summer and having forgotten
> some details since then I thought SOAPAction should accompany every
> SOAP request message. Today's discussion under this subject has lead
> me to deciding to forget about the SOAPAction header entirely because
> it's useless unless I want to have routing before processing, but even
> then I can do that using different urls.
>
> I used SOAPAction header for differentiating between instances. Using
> SOAPAction or even the URL for that purpose means showing that the
> instances are really different, but the main reason I used that over
> information in headers is that you can send _the_same_message_ to
> different addresses in order to request a service from different
> instances. Otherwise (putting the instance identifier into a header or
> even into the body) would feel like a good ole C style:
>
> method(object, param, param, param)
>
> instead of
>
> object->method(param, param, param).
>
> Thinking about it now leads me to the conclusion, though, that that's
> what headers were created for.
>
> James proposes to use the body's namespace to do the differentiation
> of services (I'll talk about instances later) but that's feasible if
> it's you who creates the service description. If you get a service
> description and want to comply to it relying only on namespaces won't
> do if the creator of that description put all the messages into a
> single namespace. So as Glen just wrote (I just received his mail on
> that) you generally have be ready to dispatch on the QName of the body
> element or more.
>
> And using namespaces to differentiate between instances is IMO
> completely wrong because this way the whole body would have to change
> when sending to a different instance. That just feels wrong (I lack
> the vocabulary at this hour).
>
> Now to the local/global handlers: why differentiate? If this
> differentiation means only either mentioning the class in a
> message-specific section of the configuration or in the fallback
> section. then that's OK. And we could even have a whole configuration
> tree where there would be classes of services and classes of classes
> etc. and the root - the last fallback.
>
> As for the xsi:type issue: what James proposes seems great to me with
> only one exception: Why not use the same solution on the client side -
> instead of only defaulting to strings (which could be useful on either
> side if no type mapping is given) the client could specify the mapping
> as well. I don't think Deployment Description should only be
> understood as a static file - an equivalent can be given to the
> client-side SOAP tools during the client set-up. 8-)
>
> BTW: defaulting to strings is what we do in our lower (non-SCL) layer
> in IdooXoap anyway.
>
> Now lemme go home. 8-)
>
> Jacek Kopecky
> Idoox s.r.o.
>
>
>
> On Tue, 5 Sep 2000, James Snell wrote:
>
> > Alrighty... good conversation... sorry I missed it... on a short break
so
> > lemme see if I can't offer some comment on the discussion.
> >
> > >From what I gathered, the basic gist of the technical conversation was
> how
> > to map Messages to Services in terms of the Service Description API.
> >
> > First of all, I absolutely concur that a standardize API is required
for
> > service descriptions, irregardless of the actual grammar that is used.
> The
> > Service Description architecture should also consist of a plugable
> approach
> > that would allow for multiple service description language processors.
> >
> > In terms of how a message gets dispatched, I was imagining a URI-based
> > approach based around the relationship of Ports/Bindings/Messages.
> >
> > Basically, a server exposes X number of ports. Each Port exposes Y
> number
> > of bindings, A binding implements Z number of Message Handlers (e.g.
"A
> > MESSAGE HANDLER is BOUND to PORT ABC").
> >
> > If you look at SCL (for example), Messages are defined, then grouped
> > according to the ports that they are bound to. The public service
> > description would include this information.
> >
> > Internally, we would keep a separate configuration file that would map
> the
> > Port Message Bindings to particular software components. Does that
make
> > sense?
> >
> > How the processor makes the leap (from receiving a message to knowing
> which
> > Port and Binding to dispatch it to) is by parsing the namespace for the
> > message. Essentially, the namespace for the payload (the message)
would
> be
> > defined as the URI of the server + the port name + the binding name +
the
> > message name:
> >
> > For example:
> >
> > http://soap.acmeco.com/portname/bindingname#messagetype
> >
> >
> > The SOAPAction header (according to the spec) is only available for the
> HTTP
> > binding. Or at least, that is the only context in which it is used...
> there
> > is no explicit statement forbidding its use in other bindings, however,
> at
> > the same time there is no explicit statement allowing its use either,
> > meaning that it should not be relied upon for any other transport
> binding.
> >
> > The only other way to uniquely identify a message and link it to a
> > particular port binding is through a unique namespace identifier.
Since
> the
> > namespace is based on the port and binding names, the server processor
> could
> > absolutely identify and route the message accordingly.
> >
> > In summary, we would have a public service description and a private
> > configuration file. The service description would instruct the client
> how
> > to build the request, including which namespace to use for the request.
> The
> > configuration file would map the namespaces to one or more message
> handlers.
> >
> >
> > The same basic technique can be used for header processing. The
headers
> can
> > be uniquely identified by namespace, with the configuration file
> providing
> > the mapping information that tells the processor how to handle it.
> >
> > Different services can map the same headers differently by employing a
> > different Namespace URI.
> >
> > For example:
> >
> > http://soap.acmeco.com/someheader/#1
> > http://soap.acmeco.com/someheader/#2
> >
> > Or something along these lines.
> >
> > Does that make any sense? I could be way off here, so if I am, just
give
> me
> > a virtual boot-to-the-head and set me straight :-)
> >
> > A couple of other comments:
> >
> > We would need a distinction between the following types of components:
> >
> > 1. Local Message Handlers
> > A Message Handler is designed to handle the specific processing
> > requirements of a single type of message bound to a single port.
> >
> > 2. Global Message Handlers
> > A Global Message Handler is a SOAP Processor plugin designed to
> handle
> > a particular type of message no matter which port it is assigned to.
> > Essentially, when a message is received and routed to a particular
port,
> the
> > processor would first look to see if a Local Message Handler exists for
> that
> > message at that port, and if not, the processor would then query the
> Global
> > Message Handler registry to see if a global message handler has been
> defined
> > for that type of message. I haven't worked out all of the details for
> this
> > yet, so please be patient. These would include support for some kind
of
> > transaction based messaging architecture, etc that would be considered
> > global for all services.
> >
> > 3. Message Specific Header Handlers
> > A Message Specific Header Handler is a header handler
specifically
> > designed to provide context to the Local Message Handlers. This would
> > include security/etc implemented for just a single service.
> >
> > 4. Global Header Handlers
> > Global Header Handlers are handlers designed to handle specific
> types
> > of headers irregardless of the message they accompany. This would
> include
> > security/etc implemented across all services, etc.
> >
> > 5. Processor Plugins
> > Processor Plugins are code components that layer in new
> functionality
> > to the processor that is completely independent of messages or headers.
> > These would include such facilities as Logging, global transforms, etc.
> >
> >
> > As for the existing xsi:type issue, why not just add the ability for a
> > developer to map each parameter in the call to a specific xsi:type
value
> and
> > store that in the deployment descriptor? When a call is received that
> does
> > not include those values, then query the deployment descriptor and
> retrieve
> > them from there. This would work great for the server side. As for
the
> > client side, I would say that if a response is received without the
> > xsi:types, default each to string values. It's not the perfect
approach,
> > but it would work. And if we are planning a rewrite based around
service
> > descriptions anyway, doesn't it make sense to punt in the short term
and
> do
> > it right the next go around? Again, I may be way off, but to me, I
think
> > that this is a valid approach to take.
> >
> > Let's see... anything else ((james is switching back and forth between
> the
> > log and the email...))
> >
> > Oh yeah.... using the Service Description API, we can transform the
> internal
> > configuration file automatically to whichever Service Description
Grammar
> is
> > needed... again, I haven't worked out all the details on this yet, but,
> in
> > theory it sounds logical.
> >
> > Oh and Glen... if you plan on going further south (Fresno area) on your
> > vacation, drop me a line. :-)
> >
> > - James
> >
>
>
>
>
============================================================================
> This message is intended only for the use of the Addressee(s) and may
> contain information that is PRIVILEGED and CONFIDENTIAL. If you are not
> the intended recipient, dissemination of this communication is prohibited.
> If you have received this communication in error, please erase all copies
> of the message and its attachments and notify postmaster@verticalnet.com
> immediately.
>
============================================================================
>
RE: Accessing instances in SOAP - ideas?
Posted by Ara Kassabian <ar...@techone.com>.
Oops! I take it back. We have NOT submitted it yet (we submitted a message
transformer framework). But we should be soon (You can talk to Vahe to find
out more in the meantime--avahe@techone.com).
----------------------------------------------
Ara Kassabian
TechOne, Inc.
100 N. Brand Blvd., Suite 605
Glendale, CA 91203
http://www.techone.com
ara@techone.com
(510) 729-6750 Ext. 396 (Oakland)
(818) 476-0055 (Los Angeles)
-----Original Message-----
From: Ara Kassabian [mailto:ara@techone.com]
Sent: Wednesday, September 06, 2000 2:06 PM
To: soap-dev@xml.apache.org; Jackson.George@necx.com;
pzf@hursley.ibm.com
Cc: Vahe Amirbekyan
Subject: RE: Accessing instances in SOAP - ideas?
We have actually implemented a locator framework to handle the situation you
are referring to. The code was submitted to the SOAP mailing list a few days
ago. This is part of our larger XML messaging strategy. If you are
interested in further details, I suggest you take a look at the code and/or
contact us directly (Note: The code was submitted by Vahe Amirbekyan
(avahe@techone.com), in case you want to do a search through the archive).
----------------------------------------------
Ara Kassabian
TechOne, Inc.
100 N. Brand Blvd., Suite 605
Glendale, CA 91203
http://www.techone.com
ara@techone.com
(510) 729-6750 Ext. 396 (Oakland)
(818) 476-0055 (Los Angeles)
-----Original Message-----
From: Paul Fremantle [mailto:pzf@hursley.ibm.com]
Sent: Wednesday, September 06, 2000 1:12 PM
To: soap-dev@xml.apache.org; Jackson.George@necx.com
Subject: Accessing instances in SOAP - ideas?
Jackson, SOAP gurus!
I've been thinking about this (how to specify instances) for EJB instances.
I've come up with two approaches. One is to call the EJB home to get the a
URI for the instance. This has the benefit of not requiring special headers.
The other is to put an instance identifier in a header. I'm interested in
comments on both approaches, and any other approaches?
I've prototyped both, and the Java Apache client code in each case is ok. I
haven't really thought through lightweight clients and other clients.
Regards,
Paul
----- Original Message -----
From: "Jackson George" <Ja...@necx.com>
To: <so...@xml.apache.org>
Sent: Wednesday, September 06, 2000 4:31 PM
Subject: Jackson : How to Write the Server Side SOAP servlet to Communicate
to the the STOCK Quote
>
>
> Hi
> We are trying to implement SOAP with Java, But for 'stockquoteClient'
,
> how to write
> a Java Servlet that could communicate to the EJB and Call a Method, What
I
> have to do for this
> Server side Servlet. I know that Using RMI-from this Servlet I could
> communicate to the EJB.
> But Anybody COULD PLEASE send me a Step by step Instructions How to set
up
> the Client
> and How set-up the SOAP server , Please send me the Code too.
>
> http://www.xmethods.com/detail.html?id=2\
>
> I am a beginner in SOAP, Just a Beginner Hope you could help me
> Appreciate , Thanks in advance
> Jackson George
> jgeorge@necx.com
>
> -----------------------------------
>
> -----Original Message-----
> From: Jacek Kopecky [mailto:jacek@idoox.com]
> Sent: Tuesday, September 05, 2000 6:39 PM
> To: 'soap-dev@xml.apache.org'
> Subject: RE: IRC log
>
>
> Just a few quick remarks before I go to sleep. 8-)
>
> I first read the SOAP spec. early this summer and having forgotten
> some details since then I thought SOAPAction should accompany every
> SOAP request message. Today's discussion under this subject has lead
> me to deciding to forget about the SOAPAction header entirely because
> it's useless unless I want to have routing before processing, but even
> then I can do that using different urls.
>
> I used SOAPAction header for differentiating between instances. Using
> SOAPAction or even the URL for that purpose means showing that the
> instances are really different, but the main reason I used that over
> information in headers is that you can send _the_same_message_ to
> different addresses in order to request a service from different
> instances. Otherwise (putting the instance identifier into a header or
> even into the body) would feel like a good ole C style:
>
> method(object, param, param, param)
>
> instead of
>
> object->method(param, param, param).
>
> Thinking about it now leads me to the conclusion, though, that that's
> what headers were created for.
>
> James proposes to use the body's namespace to do the differentiation
> of services (I'll talk about instances later) but that's feasible if
> it's you who creates the service description. If you get a service
> description and want to comply to it relying only on namespaces won't
> do if the creator of that description put all the messages into a
> single namespace. So as Glen just wrote (I just received his mail on
> that) you generally have be ready to dispatch on the QName of the body
> element or more.
>
> And using namespaces to differentiate between instances is IMO
> completely wrong because this way the whole body would have to change
> when sending to a different instance. That just feels wrong (I lack
> the vocabulary at this hour).
>
> Now to the local/global handlers: why differentiate? If this
> differentiation means only either mentioning the class in a
> message-specific section of the configuration or in the fallback
> section. then that's OK. And we could even have a whole configuration
> tree where there would be classes of services and classes of classes
> etc. and the root - the last fallback.
>
> As for the xsi:type issue: what James proposes seems great to me with
> only one exception: Why not use the same solution on the client side -
> instead of only defaulting to strings (which could be useful on either
> side if no type mapping is given) the client could specify the mapping
> as well. I don't think Deployment Description should only be
> understood as a static file - an equivalent can be given to the
> client-side SOAP tools during the client set-up. 8-)
>
> BTW: defaulting to strings is what we do in our lower (non-SCL) layer
> in IdooXoap anyway.
>
> Now lemme go home. 8-)
>
> Jacek Kopecky
> Idoox s.r.o.
>
>
>
> On Tue, 5 Sep 2000, James Snell wrote:
>
> > Alrighty... good conversation... sorry I missed it... on a short break
so
> > lemme see if I can't offer some comment on the discussion.
> >
> > >From what I gathered, the basic gist of the technical conversation was
> how
> > to map Messages to Services in terms of the Service Description API.
> >
> > First of all, I absolutely concur that a standardize API is required
for
> > service descriptions, irregardless of the actual grammar that is used.
> The
> > Service Description architecture should also consist of a plugable
> approach
> > that would allow for multiple service description language processors.
> >
> > In terms of how a message gets dispatched, I was imagining a URI-based
> > approach based around the relationship of Ports/Bindings/Messages.
> >
> > Basically, a server exposes X number of ports. Each Port exposes Y
> number
> > of bindings, A binding implements Z number of Message Handlers (e.g.
"A
> > MESSAGE HANDLER is BOUND to PORT ABC").
> >
> > If you look at SCL (for example), Messages are defined, then grouped
> > according to the ports that they are bound to. The public service
> > description would include this information.
> >
> > Internally, we would keep a separate configuration file that would map
> the
> > Port Message Bindings to particular software components. Does that
make
> > sense?
> >
> > How the processor makes the leap (from receiving a message to knowing
> which
> > Port and Binding to dispatch it to) is by parsing the namespace for the
> > message. Essentially, the namespace for the payload (the message)
would
> be
> > defined as the URI of the server + the port name + the binding name +
the
> > message name:
> >
> > For example:
> >
> > http://soap.acmeco.com/portname/bindingname#messagetype
> >
> >
> > The SOAPAction header (according to the spec) is only available for the
> HTTP
> > binding. Or at least, that is the only context in which it is used...
> there
> > is no explicit statement forbidding its use in other bindings, however,
> at
> > the same time there is no explicit statement allowing its use either,
> > meaning that it should not be relied upon for any other transport
> binding.
> >
> > The only other way to uniquely identify a message and link it to a
> > particular port binding is through a unique namespace identifier.
Since
> the
> > namespace is based on the port and binding names, the server processor
> could
> > absolutely identify and route the message accordingly.
> >
> > In summary, we would have a public service description and a private
> > configuration file. The service description would instruct the client
> how
> > to build the request, including which namespace to use for the request.
> The
> > configuration file would map the namespaces to one or more message
> handlers.
> >
> >
> > The same basic technique can be used for header processing. The
headers
> can
> > be uniquely identified by namespace, with the configuration file
> providing
> > the mapping information that tells the processor how to handle it.
> >
> > Different services can map the same headers differently by employing a
> > different Namespace URI.
> >
> > For example:
> >
> > http://soap.acmeco.com/someheader/#1
> > http://soap.acmeco.com/someheader/#2
> >
> > Or something along these lines.
> >
> > Does that make any sense? I could be way off here, so if I am, just
give
> me
> > a virtual boot-to-the-head and set me straight :-)
> >
> > A couple of other comments:
> >
> > We would need a distinction between the following types of components:
> >
> > 1. Local Message Handlers
> > A Message Handler is designed to handle the specific processing
> > requirements of a single type of message bound to a single port.
> >
> > 2. Global Message Handlers
> > A Global Message Handler is a SOAP Processor plugin designed to
> handle
> > a particular type of message no matter which port it is assigned to.
> > Essentially, when a message is received and routed to a particular
port,
> the
> > processor would first look to see if a Local Message Handler exists for
> that
> > message at that port, and if not, the processor would then query the
> Global
> > Message Handler registry to see if a global message handler has been
> defined
> > for that type of message. I haven't worked out all of the details for
> this
> > yet, so please be patient. These would include support for some kind
of
> > transaction based messaging architecture, etc that would be considered
> > global for all services.
> >
> > 3. Message Specific Header Handlers
> > A Message Specific Header Handler is a header handler
specifically
> > designed to provide context to the Local Message Handlers. This would
> > include security/etc implemented for just a single service.
> >
> > 4. Global Header Handlers
> > Global Header Handlers are handlers designed to handle specific
> types
> > of headers irregardless of the message they accompany. This would
> include
> > security/etc implemented across all services, etc.
> >
> > 5. Processor Plugins
> > Processor Plugins are code components that layer in new
> functionality
> > to the processor that is completely independent of messages or headers.
> > These would include such facilities as Logging, global transforms, etc.
> >
> >
> > As for the existing xsi:type issue, why not just add the ability for a
> > developer to map each parameter in the call to a specific xsi:type
value
> and
> > store that in the deployment descriptor? When a call is received that
> does
> > not include those values, then query the deployment descriptor and
> retrieve
> > them from there. This would work great for the server side. As for
the
> > client side, I would say that if a response is received without the
> > xsi:types, default each to string values. It's not the perfect
approach,
> > but it would work. And if we are planning a rewrite based around
service
> > descriptions anyway, doesn't it make sense to punt in the short term
and
> do
> > it right the next go around? Again, I may be way off, but to me, I
think
> > that this is a valid approach to take.
> >
> > Let's see... anything else ((james is switching back and forth between
> the
> > log and the email...))
> >
> > Oh yeah.... using the Service Description API, we can transform the
> internal
> > configuration file automatically to whichever Service Description
Grammar
> is
> > needed... again, I haven't worked out all the details on this yet, but,
> in
> > theory it sounds logical.
> >
> > Oh and Glen... if you plan on going further south (Fresno area) on your
> > vacation, drop me a line. :-)
> >
> > - James
> >
>
>
>
>
============================================================================
> This message is intended only for the use of the Addressee(s) and may
> contain information that is PRIVILEGED and CONFIDENTIAL. If you are not
> the intended recipient, dissemination of this communication is prohibited.
> If you have received this communication in error, please erase all copies
> of the message and its attachments and notify postmaster@verticalnet.com
> immediately.
>
============================================================================
>
RE: Accessing instances in SOAP - ideas?
Posted by Ara Kassabian <ar...@techone.com>.
Oops! I take it back. We have NOT submitted it yet (we submitted a message
transformer framework). But we should be soon (You can talk to Vahe to find
out more in the meantime--avahe@techone.com).
----------------------------------------------
Ara Kassabian
TechOne, Inc.
100 N. Brand Blvd., Suite 605
Glendale, CA 91203
http://www.techone.com
ara@techone.com
(510) 729-6750 Ext. 396 (Oakland)
(818) 476-0055 (Los Angeles)
-----Original Message-----
From: Ara Kassabian [mailto:ara@techone.com]
Sent: Wednesday, September 06, 2000 2:06 PM
To: soap-dev@xml.apache.org; Jackson.George@necx.com;
pzf@hursley.ibm.com
Cc: Vahe Amirbekyan
Subject: RE: Accessing instances in SOAP - ideas?
We have actually implemented a locator framework to handle the situation you
are referring to. The code was submitted to the SOAP mailing list a few days
ago. This is part of our larger XML messaging strategy. If you are
interested in further details, I suggest you take a look at the code and/or
contact us directly (Note: The code was submitted by Vahe Amirbekyan
(avahe@techone.com), in case you want to do a search through the archive).
----------------------------------------------
Ara Kassabian
TechOne, Inc.
100 N. Brand Blvd., Suite 605
Glendale, CA 91203
http://www.techone.com
ara@techone.com
(510) 729-6750 Ext. 396 (Oakland)
(818) 476-0055 (Los Angeles)
-----Original Message-----
From: Paul Fremantle [mailto:pzf@hursley.ibm.com]
Sent: Wednesday, September 06, 2000 1:12 PM
To: soap-dev@xml.apache.org; Jackson.George@necx.com
Subject: Accessing instances in SOAP - ideas?
Jackson, SOAP gurus!
I've been thinking about this (how to specify instances) for EJB instances.
I've come up with two approaches. One is to call the EJB home to get the a
URI for the instance. This has the benefit of not requiring special headers.
The other is to put an instance identifier in a header. I'm interested in
comments on both approaches, and any other approaches?
I've prototyped both, and the Java Apache client code in each case is ok. I
haven't really thought through lightweight clients and other clients.
Regards,
Paul
----- Original Message -----
From: "Jackson George" <Ja...@necx.com>
To: <so...@xml.apache.org>
Sent: Wednesday, September 06, 2000 4:31 PM
Subject: Jackson : How to Write the Server Side SOAP servlet to Communicate
to the the STOCK Quote
>
>
> Hi
> We are trying to implement SOAP with Java, But for 'stockquoteClient'
,
> how to write
> a Java Servlet that could communicate to the EJB and Call a Method, What
I
> have to do for this
> Server side Servlet. I know that Using RMI-from this Servlet I could
> communicate to the EJB.
> But Anybody COULD PLEASE send me a Step by step Instructions How to set
up
> the Client
> and How set-up the SOAP server , Please send me the Code too.
>
> http://www.xmethods.com/detail.html?id=2\
>
> I am a beginner in SOAP, Just a Beginner Hope you could help me
> Appreciate , Thanks in advance
> Jackson George
> jgeorge@necx.com
>
> -----------------------------------
>
> -----Original Message-----
> From: Jacek Kopecky [mailto:jacek@idoox.com]
> Sent: Tuesday, September 05, 2000 6:39 PM
> To: 'soap-dev@xml.apache.org'
> Subject: RE: IRC log
>
>
> Just a few quick remarks before I go to sleep. 8-)
>
> I first read the SOAP spec. early this summer and having forgotten
> some details since then I thought SOAPAction should accompany every
> SOAP request message. Today's discussion under this subject has lead
> me to deciding to forget about the SOAPAction header entirely because
> it's useless unless I want to have routing before processing, but even
> then I can do that using different urls.
>
> I used SOAPAction header for differentiating between instances. Using
> SOAPAction or even the URL for that purpose means showing that the
> instances are really different, but the main reason I used that over
> information in headers is that you can send _the_same_message_ to
> different addresses in order to request a service from different
> instances. Otherwise (putting the instance identifier into a header or
> even into the body) would feel like a good ole C style:
>
> method(object, param, param, param)
>
> instead of
>
> object->method(param, param, param).
>
> Thinking about it now leads me to the conclusion, though, that that's
> what headers were created for.
>
> James proposes to use the body's namespace to do the differentiation
> of services (I'll talk about instances later) but that's feasible if
> it's you who creates the service description. If you get a service
> description and want to comply to it relying only on namespaces won't
> do if the creator of that description put all the messages into a
> single namespace. So as Glen just wrote (I just received his mail on
> that) you generally have be ready to dispatch on the QName of the body
> element or more.
>
> And using namespaces to differentiate between instances is IMO
> completely wrong because this way the whole body would have to change
> when sending to a different instance. That just feels wrong (I lack
> the vocabulary at this hour).
>
> Now to the local/global handlers: why differentiate? If this
> differentiation means only either mentioning the class in a
> message-specific section of the configuration or in the fallback
> section. then that's OK. And we could even have a whole configuration
> tree where there would be classes of services and classes of classes
> etc. and the root - the last fallback.
>
> As for the xsi:type issue: what James proposes seems great to me with
> only one exception: Why not use the same solution on the client side -
> instead of only defaulting to strings (which could be useful on either
> side if no type mapping is given) the client could specify the mapping
> as well. I don't think Deployment Description should only be
> understood as a static file - an equivalent can be given to the
> client-side SOAP tools during the client set-up. 8-)
>
> BTW: defaulting to strings is what we do in our lower (non-SCL) layer
> in IdooXoap anyway.
>
> Now lemme go home. 8-)
>
> Jacek Kopecky
> Idoox s.r.o.
>
>
>
> On Tue, 5 Sep 2000, James Snell wrote:
>
> > Alrighty... good conversation... sorry I missed it... on a short break
so
> > lemme see if I can't offer some comment on the discussion.
> >
> > >From what I gathered, the basic gist of the technical conversation was
> how
> > to map Messages to Services in terms of the Service Description API.
> >
> > First of all, I absolutely concur that a standardize API is required
for
> > service descriptions, irregardless of the actual grammar that is used.
> The
> > Service Description architecture should also consist of a plugable
> approach
> > that would allow for multiple service description language processors.
> >
> > In terms of how a message gets dispatched, I was imagining a URI-based
> > approach based around the relationship of Ports/Bindings/Messages.
> >
> > Basically, a server exposes X number of ports. Each Port exposes Y
> number
> > of bindings, A binding implements Z number of Message Handlers (e.g.
"A
> > MESSAGE HANDLER is BOUND to PORT ABC").
> >
> > If you look at SCL (for example), Messages are defined, then grouped
> > according to the ports that they are bound to. The public service
> > description would include this information.
> >
> > Internally, we would keep a separate configuration file that would map
> the
> > Port Message Bindings to particular software components. Does that
make
> > sense?
> >
> > How the processor makes the leap (from receiving a message to knowing
> which
> > Port and Binding to dispatch it to) is by parsing the namespace for the
> > message. Essentially, the namespace for the payload (the message)
would
> be
> > defined as the URI of the server + the port name + the binding name +
the
> > message name:
> >
> > For example:
> >
> > http://soap.acmeco.com/portname/bindingname#messagetype
> >
> >
> > The SOAPAction header (according to the spec) is only available for the
> HTTP
> > binding. Or at least, that is the only context in which it is used...
> there
> > is no explicit statement forbidding its use in other bindings, however,
> at
> > the same time there is no explicit statement allowing its use either,
> > meaning that it should not be relied upon for any other transport
> binding.
> >
> > The only other way to uniquely identify a message and link it to a
> > particular port binding is through a unique namespace identifier.
Since
> the
> > namespace is based on the port and binding names, the server processor
> could
> > absolutely identify and route the message accordingly.
> >
> > In summary, we would have a public service description and a private
> > configuration file. The service description would instruct the client
> how
> > to build the request, including which namespace to use for the request.
> The
> > configuration file would map the namespaces to one or more message
> handlers.
> >
> >
> > The same basic technique can be used for header processing. The
headers
> can
> > be uniquely identified by namespace, with the configuration file
> providing
> > the mapping information that tells the processor how to handle it.
> >
> > Different services can map the same headers differently by employing a
> > different Namespace URI.
> >
> > For example:
> >
> > http://soap.acmeco.com/someheader/#1
> > http://soap.acmeco.com/someheader/#2
> >
> > Or something along these lines.
> >
> > Does that make any sense? I could be way off here, so if I am, just
give
> me
> > a virtual boot-to-the-head and set me straight :-)
> >
> > A couple of other comments:
> >
> > We would need a distinction between the following types of components:
> >
> > 1. Local Message Handlers
> > A Message Handler is designed to handle the specific processing
> > requirements of a single type of message bound to a single port.
> >
> > 2. Global Message Handlers
> > A Global Message Handler is a SOAP Processor plugin designed to
> handle
> > a particular type of message no matter which port it is assigned to.
> > Essentially, when a message is received and routed to a particular
port,
> the
> > processor would first look to see if a Local Message Handler exists for
> that
> > message at that port, and if not, the processor would then query the
> Global
> > Message Handler registry to see if a global message handler has been
> defined
> > for that type of message. I haven't worked out all of the details for
> this
> > yet, so please be patient. These would include support for some kind
of
> > transaction based messaging architecture, etc that would be considered
> > global for all services.
> >
> > 3. Message Specific Header Handlers
> > A Message Specific Header Handler is a header handler
specifically
> > designed to provide context to the Local Message Handlers. This would
> > include security/etc implemented for just a single service.
> >
> > 4. Global Header Handlers
> > Global Header Handlers are handlers designed to handle specific
> types
> > of headers irregardless of the message they accompany. This would
> include
> > security/etc implemented across all services, etc.
> >
> > 5. Processor Plugins
> > Processor Plugins are code components that layer in new
> functionality
> > to the processor that is completely independent of messages or headers.
> > These would include such facilities as Logging, global transforms, etc.
> >
> >
> > As for the existing xsi:type issue, why not just add the ability for a
> > developer to map each parameter in the call to a specific xsi:type
value
> and
> > store that in the deployment descriptor? When a call is received that
> does
> > not include those values, then query the deployment descriptor and
> retrieve
> > them from there. This would work great for the server side. As for
the
> > client side, I would say that if a response is received without the
> > xsi:types, default each to string values. It's not the perfect
approach,
> > but it would work. And if we are planning a rewrite based around
service
> > descriptions anyway, doesn't it make sense to punt in the short term
and
> do
> > it right the next go around? Again, I may be way off, but to me, I
think
> > that this is a valid approach to take.
> >
> > Let's see... anything else ((james is switching back and forth between
> the
> > log and the email...))
> >
> > Oh yeah.... using the Service Description API, we can transform the
> internal
> > configuration file automatically to whichever Service Description
Grammar
> is
> > needed... again, I haven't worked out all the details on this yet, but,
> in
> > theory it sounds logical.
> >
> > Oh and Glen... if you plan on going further south (Fresno area) on your
> > vacation, drop me a line. :-)
> >
> > - James
> >
>
>
>
>
============================================================================
> This message is intended only for the use of the Addressee(s) and may
> contain information that is PRIVILEGED and CONFIDENTIAL. If you are not
> the intended recipient, dissemination of this communication is prohibited.
> If you have received this communication in error, please erase all copies
> of the message and its attachments and notify postmaster@verticalnet.com
> immediately.
>
============================================================================
>
RE: Accessing instances in SOAP - ideas?
Posted by Ara Kassabian <ar...@techone.com>.
We have actually implemented a locator framework to handle the situation you
are referring to. The code was submitted to the SOAP mailing list a few days
ago. This is part of our larger XML messaging strategy. If you are
interested in further details, I suggest you take a look at the code and/or
contact us directly (Note: The code was submitted by Vahe Amirbekyan
(avahe@techone.com), in case you want to do a search through the archive).
----------------------------------------------
Ara Kassabian
TechOne, Inc.
100 N. Brand Blvd., Suite 605
Glendale, CA 91203
http://www.techone.com
ara@techone.com
(510) 729-6750 Ext. 396 (Oakland)
(818) 476-0055 (Los Angeles)
-----Original Message-----
From: Paul Fremantle [mailto:pzf@hursley.ibm.com]
Sent: Wednesday, September 06, 2000 1:12 PM
To: soap-dev@xml.apache.org; Jackson.George@necx.com
Subject: Accessing instances in SOAP - ideas?
Jackson, SOAP gurus!
I've been thinking about this (how to specify instances) for EJB instances.
I've come up with two approaches. One is to call the EJB home to get the a
URI for the instance. This has the benefit of not requiring special headers.
The other is to put an instance identifier in a header. I'm interested in
comments on both approaches, and any other approaches?
I've prototyped both, and the Java Apache client code in each case is ok. I
haven't really thought through lightweight clients and other clients.
Regards,
Paul
----- Original Message -----
From: "Jackson George" <Ja...@necx.com>
To: <so...@xml.apache.org>
Sent: Wednesday, September 06, 2000 4:31 PM
Subject: Jackson : How to Write the Server Side SOAP servlet to Communicate
to the the STOCK Quote
>
>
> Hi
> We are trying to implement SOAP with Java, But for 'stockquoteClient'
,
> how to write
> a Java Servlet that could communicate to the EJB and Call a Method, What
I
> have to do for this
> Server side Servlet. I know that Using RMI-from this Servlet I could
> communicate to the EJB.
> But Anybody COULD PLEASE send me a Step by step Instructions How to set
up
> the Client
> and How set-up the SOAP server , Please send me the Code too.
>
> http://www.xmethods.com/detail.html?id=2\
>
> I am a beginner in SOAP, Just a Beginner Hope you could help me
> Appreciate , Thanks in advance
> Jackson George
> jgeorge@necx.com
>
> -----------------------------------
>
> -----Original Message-----
> From: Jacek Kopecky [mailto:jacek@idoox.com]
> Sent: Tuesday, September 05, 2000 6:39 PM
> To: 'soap-dev@xml.apache.org'
> Subject: RE: IRC log
>
>
> Just a few quick remarks before I go to sleep. 8-)
>
> I first read the SOAP spec. early this summer and having forgotten
> some details since then I thought SOAPAction should accompany every
> SOAP request message. Today's discussion under this subject has lead
> me to deciding to forget about the SOAPAction header entirely because
> it's useless unless I want to have routing before processing, but even
> then I can do that using different urls.
>
> I used SOAPAction header for differentiating between instances. Using
> SOAPAction or even the URL for that purpose means showing that the
> instances are really different, but the main reason I used that over
> information in headers is that you can send _the_same_message_ to
> different addresses in order to request a service from different
> instances. Otherwise (putting the instance identifier into a header or
> even into the body) would feel like a good ole C style:
>
> method(object, param, param, param)
>
> instead of
>
> object->method(param, param, param).
>
> Thinking about it now leads me to the conclusion, though, that that's
> what headers were created for.
>
> James proposes to use the body's namespace to do the differentiation
> of services (I'll talk about instances later) but that's feasible if
> it's you who creates the service description. If you get a service
> description and want to comply to it relying only on namespaces won't
> do if the creator of that description put all the messages into a
> single namespace. So as Glen just wrote (I just received his mail on
> that) you generally have be ready to dispatch on the QName of the body
> element or more.
>
> And using namespaces to differentiate between instances is IMO
> completely wrong because this way the whole body would have to change
> when sending to a different instance. That just feels wrong (I lack
> the vocabulary at this hour).
>
> Now to the local/global handlers: why differentiate? If this
> differentiation means only either mentioning the class in a
> message-specific section of the configuration or in the fallback
> section. then that's OK. And we could even have a whole configuration
> tree where there would be classes of services and classes of classes
> etc. and the root - the last fallback.
>
> As for the xsi:type issue: what James proposes seems great to me with
> only one exception: Why not use the same solution on the client side -
> instead of only defaulting to strings (which could be useful on either
> side if no type mapping is given) the client could specify the mapping
> as well. I don't think Deployment Description should only be
> understood as a static file - an equivalent can be given to the
> client-side SOAP tools during the client set-up. 8-)
>
> BTW: defaulting to strings is what we do in our lower (non-SCL) layer
> in IdooXoap anyway.
>
> Now lemme go home. 8-)
>
> Jacek Kopecky
> Idoox s.r.o.
>
>
>
> On Tue, 5 Sep 2000, James Snell wrote:
>
> > Alrighty... good conversation... sorry I missed it... on a short break
so
> > lemme see if I can't offer some comment on the discussion.
> >
> > >From what I gathered, the basic gist of the technical conversation was
> how
> > to map Messages to Services in terms of the Service Description API.
> >
> > First of all, I absolutely concur that a standardize API is required
for
> > service descriptions, irregardless of the actual grammar that is used.
> The
> > Service Description architecture should also consist of a plugable
> approach
> > that would allow for multiple service description language processors.
> >
> > In terms of how a message gets dispatched, I was imagining a URI-based
> > approach based around the relationship of Ports/Bindings/Messages.
> >
> > Basically, a server exposes X number of ports. Each Port exposes Y
> number
> > of bindings, A binding implements Z number of Message Handlers (e.g.
"A
> > MESSAGE HANDLER is BOUND to PORT ABC").
> >
> > If you look at SCL (for example), Messages are defined, then grouped
> > according to the ports that they are bound to. The public service
> > description would include this information.
> >
> > Internally, we would keep a separate configuration file that would map
> the
> > Port Message Bindings to particular software components. Does that
make
> > sense?
> >
> > How the processor makes the leap (from receiving a message to knowing
> which
> > Port and Binding to dispatch it to) is by parsing the namespace for the
> > message. Essentially, the namespace for the payload (the message)
would
> be
> > defined as the URI of the server + the port name + the binding name +
the
> > message name:
> >
> > For example:
> >
> > http://soap.acmeco.com/portname/bindingname#messagetype
> >
> >
> > The SOAPAction header (according to the spec) is only available for the
> HTTP
> > binding. Or at least, that is the only context in which it is used...
> there
> > is no explicit statement forbidding its use in other bindings, however,
> at
> > the same time there is no explicit statement allowing its use either,
> > meaning that it should not be relied upon for any other transport
> binding.
> >
> > The only other way to uniquely identify a message and link it to a
> > particular port binding is through a unique namespace identifier.
Since
> the
> > namespace is based on the port and binding names, the server processor
> could
> > absolutely identify and route the message accordingly.
> >
> > In summary, we would have a public service description and a private
> > configuration file. The service description would instruct the client
> how
> > to build the request, including which namespace to use for the request.
> The
> > configuration file would map the namespaces to one or more message
> handlers.
> >
> >
> > The same basic technique can be used for header processing. The
headers
> can
> > be uniquely identified by namespace, with the configuration file
> providing
> > the mapping information that tells the processor how to handle it.
> >
> > Different services can map the same headers differently by employing a
> > different Namespace URI.
> >
> > For example:
> >
> > http://soap.acmeco.com/someheader/#1
> > http://soap.acmeco.com/someheader/#2
> >
> > Or something along these lines.
> >
> > Does that make any sense? I could be way off here, so if I am, just
give
> me
> > a virtual boot-to-the-head and set me straight :-)
> >
> > A couple of other comments:
> >
> > We would need a distinction between the following types of components:
> >
> > 1. Local Message Handlers
> > A Message Handler is designed to handle the specific processing
> > requirements of a single type of message bound to a single port.
> >
> > 2. Global Message Handlers
> > A Global Message Handler is a SOAP Processor plugin designed to
> handle
> > a particular type of message no matter which port it is assigned to.
> > Essentially, when a message is received and routed to a particular
port,
> the
> > processor would first look to see if a Local Message Handler exists for
> that
> > message at that port, and if not, the processor would then query the
> Global
> > Message Handler registry to see if a global message handler has been
> defined
> > for that type of message. I haven't worked out all of the details for
> this
> > yet, so please be patient. These would include support for some kind
of
> > transaction based messaging architecture, etc that would be considered
> > global for all services.
> >
> > 3. Message Specific Header Handlers
> > A Message Specific Header Handler is a header handler
specifically
> > designed to provide context to the Local Message Handlers. This would
> > include security/etc implemented for just a single service.
> >
> > 4. Global Header Handlers
> > Global Header Handlers are handlers designed to handle specific
> types
> > of headers irregardless of the message they accompany. This would
> include
> > security/etc implemented across all services, etc.
> >
> > 5. Processor Plugins
> > Processor Plugins are code components that layer in new
> functionality
> > to the processor that is completely independent of messages or headers.
> > These would include such facilities as Logging, global transforms, etc.
> >
> >
> > As for the existing xsi:type issue, why not just add the ability for a
> > developer to map each parameter in the call to a specific xsi:type
value
> and
> > store that in the deployment descriptor? When a call is received that
> does
> > not include those values, then query the deployment descriptor and
> retrieve
> > them from there. This would work great for the server side. As for
the
> > client side, I would say that if a response is received without the
> > xsi:types, default each to string values. It's not the perfect
approach,
> > but it would work. And if we are planning a rewrite based around
service
> > descriptions anyway, doesn't it make sense to punt in the short term
and
> do
> > it right the next go around? Again, I may be way off, but to me, I
think
> > that this is a valid approach to take.
> >
> > Let's see... anything else ((james is switching back and forth between
> the
> > log and the email...))
> >
> > Oh yeah.... using the Service Description API, we can transform the
> internal
> > configuration file automatically to whichever Service Description
Grammar
> is
> > needed... again, I haven't worked out all the details on this yet, but,
> in
> > theory it sounds logical.
> >
> > Oh and Glen... if you plan on going further south (Fresno area) on your
> > vacation, drop me a line. :-)
> >
> > - James
> >
>
>
>
>
============================================================================
> This message is intended only for the use of the Addressee(s) and may
> contain information that is PRIVILEGED and CONFIDENTIAL. If you are not
> the intended recipient, dissemination of this communication is prohibited.
> If you have received this communication in error, please erase all copies
> of the message and its attachments and notify postmaster@verticalnet.com
> immediately.
>
============================================================================
>
RE: Accessing instances in SOAP - ideas?
Posted by Ara Kassabian <ar...@techone.com>.
We have actually implemented a locator framework to handle the situation you
are referring to. The code was submitted to the SOAP mailing list a few days
ago. This is part of our larger XML messaging strategy. If you are
interested in further details, I suggest you take a look at the code and/or
contact us directly (Note: The code was submitted by Vahe Amirbekyan
(avahe@techone.com), in case you want to do a search through the archive).
----------------------------------------------
Ara Kassabian
TechOne, Inc.
100 N. Brand Blvd., Suite 605
Glendale, CA 91203
http://www.techone.com
ara@techone.com
(510) 729-6750 Ext. 396 (Oakland)
(818) 476-0055 (Los Angeles)
-----Original Message-----
From: Paul Fremantle [mailto:pzf@hursley.ibm.com]
Sent: Wednesday, September 06, 2000 1:12 PM
To: soap-dev@xml.apache.org; Jackson.George@necx.com
Subject: Accessing instances in SOAP - ideas?
Jackson, SOAP gurus!
I've been thinking about this (how to specify instances) for EJB instances.
I've come up with two approaches. One is to call the EJB home to get the a
URI for the instance. This has the benefit of not requiring special headers.
The other is to put an instance identifier in a header. I'm interested in
comments on both approaches, and any other approaches?
I've prototyped both, and the Java Apache client code in each case is ok. I
haven't really thought through lightweight clients and other clients.
Regards,
Paul
----- Original Message -----
From: "Jackson George" <Ja...@necx.com>
To: <so...@xml.apache.org>
Sent: Wednesday, September 06, 2000 4:31 PM
Subject: Jackson : How to Write the Server Side SOAP servlet to Communicate
to the the STOCK Quote
>
>
> Hi
> We are trying to implement SOAP with Java, But for 'stockquoteClient'
,
> how to write
> a Java Servlet that could communicate to the EJB and Call a Method, What
I
> have to do for this
> Server side Servlet. I know that Using RMI-from this Servlet I could
> communicate to the EJB.
> But Anybody COULD PLEASE send me a Step by step Instructions How to set
up
> the Client
> and How set-up the SOAP server , Please send me the Code too.
>
> http://www.xmethods.com/detail.html?id=2\
>
> I am a beginner in SOAP, Just a Beginner Hope you could help me
> Appreciate , Thanks in advance
> Jackson George
> jgeorge@necx.com
>
> -----------------------------------
>
> -----Original Message-----
> From: Jacek Kopecky [mailto:jacek@idoox.com]
> Sent: Tuesday, September 05, 2000 6:39 PM
> To: 'soap-dev@xml.apache.org'
> Subject: RE: IRC log
>
>
> Just a few quick remarks before I go to sleep. 8-)
>
> I first read the SOAP spec. early this summer and having forgotten
> some details since then I thought SOAPAction should accompany every
> SOAP request message. Today's discussion under this subject has lead
> me to deciding to forget about the SOAPAction header entirely because
> it's useless unless I want to have routing before processing, but even
> then I can do that using different urls.
>
> I used SOAPAction header for differentiating between instances. Using
> SOAPAction or even the URL for that purpose means showing that the
> instances are really different, but the main reason I used that over
> information in headers is that you can send _the_same_message_ to
> different addresses in order to request a service from different
> instances. Otherwise (putting the instance identifier into a header or
> even into the body) would feel like a good ole C style:
>
> method(object, param, param, param)
>
> instead of
>
> object->method(param, param, param).
>
> Thinking about it now leads me to the conclusion, though, that that's
> what headers were created for.
>
> James proposes to use the body's namespace to do the differentiation
> of services (I'll talk about instances later) but that's feasible if
> it's you who creates the service description. If you get a service
> description and want to comply to it relying only on namespaces won't
> do if the creator of that description put all the messages into a
> single namespace. So as Glen just wrote (I just received his mail on
> that) you generally have be ready to dispatch on the QName of the body
> element or more.
>
> And using namespaces to differentiate between instances is IMO
> completely wrong because this way the whole body would have to change
> when sending to a different instance. That just feels wrong (I lack
> the vocabulary at this hour).
>
> Now to the local/global handlers: why differentiate? If this
> differentiation means only either mentioning the class in a
> message-specific section of the configuration or in the fallback
> section. then that's OK. And we could even have a whole configuration
> tree where there would be classes of services and classes of classes
> etc. and the root - the last fallback.
>
> As for the xsi:type issue: what James proposes seems great to me with
> only one exception: Why not use the same solution on the client side -
> instead of only defaulting to strings (which could be useful on either
> side if no type mapping is given) the client could specify the mapping
> as well. I don't think Deployment Description should only be
> understood as a static file - an equivalent can be given to the
> client-side SOAP tools during the client set-up. 8-)
>
> BTW: defaulting to strings is what we do in our lower (non-SCL) layer
> in IdooXoap anyway.
>
> Now lemme go home. 8-)
>
> Jacek Kopecky
> Idoox s.r.o.
>
>
>
> On Tue, 5 Sep 2000, James Snell wrote:
>
> > Alrighty... good conversation... sorry I missed it... on a short break
so
> > lemme see if I can't offer some comment on the discussion.
> >
> > >From what I gathered, the basic gist of the technical conversation was
> how
> > to map Messages to Services in terms of the Service Description API.
> >
> > First of all, I absolutely concur that a standardize API is required
for
> > service descriptions, irregardless of the actual grammar that is used.
> The
> > Service Description architecture should also consist of a plugable
> approach
> > that would allow for multiple service description language processors.
> >
> > In terms of how a message gets dispatched, I was imagining a URI-based
> > approach based around the relationship of Ports/Bindings/Messages.
> >
> > Basically, a server exposes X number of ports. Each Port exposes Y
> number
> > of bindings, A binding implements Z number of Message Handlers (e.g.
"A
> > MESSAGE HANDLER is BOUND to PORT ABC").
> >
> > If you look at SCL (for example), Messages are defined, then grouped
> > according to the ports that they are bound to. The public service
> > description would include this information.
> >
> > Internally, we would keep a separate configuration file that would map
> the
> > Port Message Bindings to particular software components. Does that
make
> > sense?
> >
> > How the processor makes the leap (from receiving a message to knowing
> which
> > Port and Binding to dispatch it to) is by parsing the namespace for the
> > message. Essentially, the namespace for the payload (the message)
would
> be
> > defined as the URI of the server + the port name + the binding name +
the
> > message name:
> >
> > For example:
> >
> > http://soap.acmeco.com/portname/bindingname#messagetype
> >
> >
> > The SOAPAction header (according to the spec) is only available for the
> HTTP
> > binding. Or at least, that is the only context in which it is used...
> there
> > is no explicit statement forbidding its use in other bindings, however,
> at
> > the same time there is no explicit statement allowing its use either,
> > meaning that it should not be relied upon for any other transport
> binding.
> >
> > The only other way to uniquely identify a message and link it to a
> > particular port binding is through a unique namespace identifier.
Since
> the
> > namespace is based on the port and binding names, the server processor
> could
> > absolutely identify and route the message accordingly.
> >
> > In summary, we would have a public service description and a private
> > configuration file. The service description would instruct the client
> how
> > to build the request, including which namespace to use for the request.
> The
> > configuration file would map the namespaces to one or more message
> handlers.
> >
> >
> > The same basic technique can be used for header processing. The
headers
> can
> > be uniquely identified by namespace, with the configuration file
> providing
> > the mapping information that tells the processor how to handle it.
> >
> > Different services can map the same headers differently by employing a
> > different Namespace URI.
> >
> > For example:
> >
> > http://soap.acmeco.com/someheader/#1
> > http://soap.acmeco.com/someheader/#2
> >
> > Or something along these lines.
> >
> > Does that make any sense? I could be way off here, so if I am, just
give
> me
> > a virtual boot-to-the-head and set me straight :-)
> >
> > A couple of other comments:
> >
> > We would need a distinction between the following types of components:
> >
> > 1. Local Message Handlers
> > A Message Handler is designed to handle the specific processing
> > requirements of a single type of message bound to a single port.
> >
> > 2. Global Message Handlers
> > A Global Message Handler is a SOAP Processor plugin designed to
> handle
> > a particular type of message no matter which port it is assigned to.
> > Essentially, when a message is received and routed to a particular
port,
> the
> > processor would first look to see if a Local Message Handler exists for
> that
> > message at that port, and if not, the processor would then query the
> Global
> > Message Handler registry to see if a global message handler has been
> defined
> > for that type of message. I haven't worked out all of the details for
> this
> > yet, so please be patient. These would include support for some kind
of
> > transaction based messaging architecture, etc that would be considered
> > global for all services.
> >
> > 3. Message Specific Header Handlers
> > A Message Specific Header Handler is a header handler
specifically
> > designed to provide context to the Local Message Handlers. This would
> > include security/etc implemented for just a single service.
> >
> > 4. Global Header Handlers
> > Global Header Handlers are handlers designed to handle specific
> types
> > of headers irregardless of the message they accompany. This would
> include
> > security/etc implemented across all services, etc.
> >
> > 5. Processor Plugins
> > Processor Plugins are code components that layer in new
> functionality
> > to the processor that is completely independent of messages or headers.
> > These would include such facilities as Logging, global transforms, etc.
> >
> >
> > As for the existing xsi:type issue, why not just add the ability for a
> > developer to map each parameter in the call to a specific xsi:type
value
> and
> > store that in the deployment descriptor? When a call is received that
> does
> > not include those values, then query the deployment descriptor and
> retrieve
> > them from there. This would work great for the server side. As for
the
> > client side, I would say that if a response is received without the
> > xsi:types, default each to string values. It's not the perfect
approach,
> > but it would work. And if we are planning a rewrite based around
service
> > descriptions anyway, doesn't it make sense to punt in the short term
and
> do
> > it right the next go around? Again, I may be way off, but to me, I
think
> > that this is a valid approach to take.
> >
> > Let's see... anything else ((james is switching back and forth between
> the
> > log and the email...))
> >
> > Oh yeah.... using the Service Description API, we can transform the
> internal
> > configuration file automatically to whichever Service Description
Grammar
> is
> > needed... again, I haven't worked out all the details on this yet, but,
> in
> > theory it sounds logical.
> >
> > Oh and Glen... if you plan on going further south (Fresno area) on your
> > vacation, drop me a line. :-)
> >
> > - James
> >
>
>
>
>
============================================================================
> This message is intended only for the use of the Addressee(s) and may
> contain information that is PRIVILEGED and CONFIDENTIAL. If you are not
> the intended recipient, dissemination of this communication is prohibited.
> If you have received this communication in error, please erase all copies
> of the message and its attachments and notify postmaster@verticalnet.com
> immediately.
>
============================================================================
>