Re: [C++] Do we need Tuscany-model.config?
Pete Robbins wrote: of course this supposes that all the schema and wsdl is local. We probably need to support the case where the wsdl is remote e.g. http://mySite/flobber.wsdl Cheers, Here's a simple way to support remote WSDLs, no need to invent our own config file for that. As a user you simply create a local WSDL containing the following: wsdl:definition ... wsdl.import namespace=http://myns/; location=http://mySite/flobber.wsdl/ ... other WSDL declarations /wsdl:definition If we just support this as a first step (and we have to support all wsdl:import schemes anyway), we provide the necessary function without requiring users to learn a new XML config language. -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [C++] Do we need Tuscany-model.config?
Pete Robbins wrote: Actually the wsdl/xsd are loaded for a particular composite so if there were more say 2 .composites in a folder and 3 xsds we would load all 3 xsds twice, once for each composite. Concrete scenarios will help us understand this better, but I'm starting to think that WSDLs are not loaded just per composite (or per composite folder). Let me illustrate this through my favorite bigbank scenario :) - The bigbank.account composite exposes an Account Service, described by bigbank.account/AccountService.wsdl. AccountService.wsdl is most naturally packaged in the bigbank.account composite folder or archive. - The system administrator creates a bigbank composite and inside it an instance of bigbank.account (a component with an implementation.composite name=bigbank.account/). - The system administrator then decides to expose the AccountService to remote Web Services clients, at a configured endpoint, with the following: service... interface.wsdl name=... Account/ binding.ws ... uri=http://myserver/account/ /service In this scenario you don't want to repeat/copy the WSDL in bigbank.composite just to override the endpoint exposing bigbank.account, so AccountService.wsdl needs to be known outside of the composite that initially packaged it. -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [C++] Do we need Tuscany-model.config?
so do you think that wsdl and shema should be loaded on a per system basis, i.e. we load all .wsdl and .xsd in either the root directory or any we find from the root down? On 15/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: Pete Robbins wrote: Actually the wsdl/xsd are loaded for a particular composite so if there were more say 2 .composites in a folder and 3 xsds we would load all 3 xsds twice, once for each composite. Concrete scenarios will help us understand this better, but I'm starting to think that WSDLs are not loaded just per composite (or per composite folder). Let me illustrate this through my favorite bigbank scenario :) - The bigbank.account composite exposes an Account Service, described by bigbank.account/AccountService.wsdl. AccountService.wsdl is most naturally packaged in the bigbank.account composite folder or archive. - The system administrator creates a bigbank composite and inside it an instance of bigbank.account (a component with an implementation.composite name=bigbank.account/). - The system administrator then decides to expose the AccountService to remote Web Services clients, at a configured endpoint, with the following: service... interface.wsdl name=... Account/ binding.ws ... uri=http://myserver/account/ /service In this scenario you don't want to repeat/copy the WSDL in bigbank.composite just to override the endpoint exposing bigbank.account, so AccountService.wsdl needs to be known outside of the composite that initially packaged it. -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] -- Pete
Re: [C++] Do we need Tuscany-model.config?
Pete Robbins wrote: so do you think that wsdl and shema should be loaded on a per system basis, i.e. we load all .wsdl and .xsd in either the root directory or any we find from the root down? Yes :) -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [C++] Do we need Tuscany-model.config?
OK! let's do it then. We'll probably need to think what the getDataFactory() method of ComponentContext/CompositeContext returns. We may want a separate instance of the datafactory for each composite so we load all the wsdl/xsd into a DataFactory which we clone for each composite. First step is to have one DF for the system. (I'm just a tad worried about DO's getting freed up... maybe I shouldn't worry ;-) ) Cheers, On 15/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: Pete Robbins wrote: so do you think that wsdl and shema should be loaded on a per system basis, i.e. we load all .wsdl and .xsd in either the root directory or any we find from the root down? Yes :) -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] -- Pete
Re: [C++] Do we need Tuscany-model.config?
This is similar to what Java had. I think htis is replaced by include.wsdl and include.xsd or soemthing. Not sure if this got into the spec. Cheers, On 11/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: It looks like Tuscany-model.config just lists the WSDLs and XSDs used in an application. Do we still need it? or could we just figure out ourselves what the WSDLs and XSDs are? Just trying to make it simpler for users to write SCA applications, and limit the number of things that they have to worry about... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] -- Pete
Re: [C++] Do we need Tuscany-model.config?
Pete Robbins wrote: This is similar to what Java had. I think htis is replaced by include.wsdl and include.xsd or soemthing. Not sure if this got into the spec. Cheers, On 11/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: It looks like Tuscany-model.config just lists the WSDLs and XSDs used in an application. Do we still need it? or could we just figure out ourselves what the WSDLs and XSDs are? Just trying to make it simpler for users to write SCA applications, and limit the number of things that they have to worry about... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] As far as I know include.wsdl and include.xsd are not in the spec yet. I'm actually thinking that placing the WSDL and XSD files in the directory containing the application should be enough to indicate that they should be used. In other words, no need to edit an XML file telling the runtime that some WSDL and XSD files are there, the runtime should be smart enough to just scan and find them... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [C++] Do we need Tuscany-model.config?
Actually the wsdl/xsd are loaded for a particular composite so if there were more say 2 .composites in a folder and 3 xsds we would load all 3 xsds twice, once for each composite. On 11/08/06, Pete Robbins [EMAIL PROTECTED] wrote: so we load any wsdl/xsd in the same folder as the .composite file? Sounds reasonable. Cheers, On 11/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: Pete Robbins wrote: This is similar to what Java had. I think htis is replaced by include.wsdl and include.xsd or soemthing. Not sure if this got into the spec. Cheers, On 11/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: It looks like Tuscany-model.config just lists the WSDLs and XSDs used in an application. Do we still need it? or could we just figure out ourselves what the WSDLs and XSDs are? Just trying to make it simpler for users to write SCA applications, and limit the number of things that they have to worry about... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] As far as I know include.wsdl and include.xsd are not in the spec yet. I'm actually thinking that placing the WSDL and XSD files in the directory containing the application should be enough to indicate that they should be used. In other words, no need to edit an XML file telling the runtime that some WSDL and XSD files are there, the runtime should be smart enough to just scan and find them... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] -- Pete -- Pete
Re: [C++] Do we need Tuscany-model.config?
of course this supposes that all the schema and wsdl is local. We probably need to support the case where the wsdl is remote e.g. http://mySite/flobber.wsdl Cheers, On 11/08/06, Pete Robbins [EMAIL PROTECTED] wrote: Actually the wsdl/xsd are loaded for a particular composite so if there were more say 2 .composites in a folder and 3 xsds we would load all 3 xsds twice, once for each composite. On 11/08/06, Pete Robbins [EMAIL PROTECTED] wrote: so we load any wsdl/xsd in the same folder as the .composite file? Sounds reasonable. Cheers, On 11/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: Pete Robbins wrote: This is similar to what Java had. I think htis is replaced by include.wsdl and include.xsd or soemthing. Not sure if this got into the spec. Cheers, On 11/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: It looks like Tuscany-model.config just lists the WSDLs and XSDs used in an application. Do we still need it? or could we just figure out ourselves what the WSDLs and XSDs are? Just trying to make it simpler for users to write SCA applications, and limit the number of things that they have to worry about... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] As far as I know include.wsdl and include.xsd are not in the spec yet. I'm actually thinking that placing the WSDL and XSD files in the directory containing the application should be enough to indicate that they should be used. In other words, no need to edit an XML file telling the runtime that some WSDL and XSD files are there, the runtime should be smart enough to just scan and find them... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] -- Pete -- Pete -- Pete
Re: [C++] Do we need Tuscany-model.config?
In the Java version we are using WSDL2.0's wsdlLocation feature: http://www.w3.org/TR/wsdl20/#wsdlLocation-aii For an example of how a service would look you might check this thread: http://mail-archives.apache.org/mod_mbox/ws-tuscany-dev/200608.mbox/% [EMAIL PROTECTED] -- Jeremy On Aug 11, 2006, at 12:57 AM, Pete Robbins wrote: of course this supposes that all the schema and wsdl is local. We probably need to support the case where the wsdl is remote e.g. http://mySite/flobber.wsdl Cheers, On 11/08/06, Pete Robbins [EMAIL PROTECTED] wrote: Actually the wsdl/xsd are loaded for a particular composite so if there were more say 2 .composites in a folder and 3 xsds we would load all 3 xsds twice, once for each composite. On 11/08/06, Pete Robbins [EMAIL PROTECTED] wrote: so we load any wsdl/xsd in the same folder as the .composite file? Sounds reasonable. Cheers, On 11/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: Pete Robbins wrote: This is similar to what Java had. I think htis is replaced by include.wsdl and include.xsd or soemthing. Not sure if this got into the spec. Cheers, On 11/08/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: It looks like Tuscany-model.config just lists the WSDLs and XSDs used in an application. Do we still need it? or could we just figure out ourselves what the WSDLs and XSDs are? Just trying to make it simpler for users to write SCA applications, and limit the number of things that they have to worry about... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: tuscany-dev- [EMAIL PROTECTED] As far as I know include.wsdl and include.xsd are not in the spec yet. I'm actually thinking that placing the WSDL and XSD files in the directory containing the application should be enough to indicate that they should be used. In other words, no need to edit an XML file telling the runtime that some WSDL and XSD files are there, the runtime should be smart enough to just scan and find them... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] -- Pete -- Pete -- Pete - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [C++] Do we need Tuscany-model.config?
I really like that auto discovery of WSDL without having to explicitly have anything pointing to it in the scdl. Didn't the really old Java code that was first given to Apache work like this? I think Ken has also mentioned that the JAXWS binding he's doing will work like that as well. Probably there still needs to be a way to explicitly define this in the SCDL, i did like the old import.wsdl way that the Java M1 code had, it was much nicer than the tuscany-model.config we used to use and i think it makes things clearer than having a wsdlLocation on the interface.wsdl or biniding.wselements. Whatever we do we should try to keep this consistent across the Java and C++ runtimes. ...ant On 8/11/06, Jean-Sebastien Delfino [EMAIL PROTECTED] wrote: snip/ As far as I know include.wsdl and include.xsd are not in the spec yet. I'm actually thinking that placing the WSDL and XSD files in the directory containing the application should be enough to indicate that they should be used. In other words, no need to edit an XML file telling the runtime that some WSDL and XSD files are there, the runtime should be smart enough to just scan and find them... -- Jean-Sebastien - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]