Technology Choices: XML-RPC vs SOAP

[Robert Olson]

>Python 2.x (I think it was 2.0 then, it's probably true that we've moved
>to 2.2)

Yup.

>The conclusion being, we should probably be targeting SOAP as our first
>web services interface, for interop, tools availability, and to remain
>true to our claim that "we adhere to standards" (not like the big bad
>you know who).  I know this is a shift, but I know way back when, I was
>also advocating building our system independent of the network interface
>so we could easily adapt network interfaces (I know there's some of this
>in Bob's venues stuff now), which was based on the pain of ripping corba
>out of voyager. If we've done that, it shouldn't be too painful to
>gracefully transition to SOAP from XML-RPC, as our primary target.

I agree that we should target SOAP. However, Not Yet. It took a fair bit of 
engineering work to make the underlying services library happy in its 
handling of XMLRPC and integration with Globus. If we put off any more 
prototyping effort now until we have a SOAP-based infrastructure we won't 
have anything to show at SC.

The other development that makes me want to wait is the work that Keith 
Jackson (pyGlobus guy, from LBL) is planning to do with respect to a native 
Python binding of the OGSA specificiation. If that work happens (I've 
pinged him about its status) we could conceivably leverage all its 
engineering work and be OGSA-compliant.

My approach in prototyping is to use good engineering practices and 
identify and isolate the network interface parts of the services. Right 
now, they're exposed via the XMLRPC service interface (via the service 
interface definition file). Later, they could be exposed via an OGSA interface.

--bob