W3C home > Mailing lists > Public > xml-dist-app@w3.org > December 2000

Re: [DS4] Remote Procedure Call (RPC)

From: Paul Denning <pauld@mitre.org>
Date: Thu, 21 Dec 2000 16:42:09 -0500
Message-Id: <5.0.0.25.0.20001221140656.0405cc90@mailsrv1.mitre.org>
To: john_ibbotson@uk.ibm.com, xml-dist-app@w3.org
At 09:29 AM 2000-12-21, john_ibbotson@uk.ibm.com wrote:
>modeled as a list of arguments, and response data modeled as a function
>result plus optional "out" parameters (or error).

I like the distinction between "response data" and "out parameters".  I 
think this is important for mapping between CORBA IDL and XP.  The response 
data could also be optional, as in (void) foo (...).

I don't think it is important to have XP define the equivalent of CORBA IDL 
"inout" parameters.  "in" and "out" seem sufficient, since XP is mostly 
concerned with the wire as opposed to the interface.

Do we need separate DS's for the case where the Ultimate XP Receiver 
converts the XP RPC request into a CORBA object call or EJB 
invocation.  The Ultimate XP Receiver, in this case, would wait for the 
CORBA or EJB response before returning the XP response.  I think we need to 
then distinguish between XP protocol errors and "back end" errors or 
exceptions.  We may want to suppress the "back end" exception at the 
Ultimate XP Receiver and not return the exception to the Original XP 
Sender.  Or we may want a way to convey the information from a "back end" 
exception back to the XP Sender.  This would be different than if the 
Ultimate XP Receiver receives the RPC request, but found a problem before 
it even invoked the back end.  XP error handling needs to accommodate both 
cases.

Paul
Received on Thursday, 21 December 2000 16:44:00 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:58:58 GMT