[wsrp-interfaces] 8/15 Conference call summary

[Michael Freedman <Michael.Freedman@oracle.com>]

On the call today: Mike Freedman, Rich Thompson, Alejandro Abdelnur, Yossi
Tamari, Carsten Leue, Joe Rudnicki, Charlie Wiecha, Tim Jones, Alan Kropp,
and Shankar Ramaswamy.

Summary:
Charlie asked that we schedule time to talk soon about the property proposal sent to the lists yesterday.  We decided to try and include this in either next Tuesdays or Thursdays discussions.  Please read the proposal and send initial comments to the discussion lists.

The bulk of the meeting discussed interface factoring.  Primary discussion concerned educating ourselves on impacts to the the consumer with regards to handling multiple factors.  Carsten provided an overview indicating that consumers would tend to isolate the impact of factoring by providing a consumer side uniform interface with which it interacts.  The implementation of this interface would be a dynamic composition of the necessary factors supported by the producer.

From this overview we moved into a technical discussion concerning how a consumer determines what factors or interfaces a producer specifically supports.  We determined that WSRP/WSIA will have to define WSDL specific names for either port types, bindings, or ports. Having a constant name one can look up in the WSDL that represents a factor allows a consumer to easily determine factors.  We cut the discussion off before defining which WSDL element(s) we would have to define standard/predefined names for.
 

> Note: After the concall I did some research and had a conversation
> with a colleague. I suspect we will want to define standard names for both
> bindings and for port types but not ports.  Though JAX-RPC's consumer
> API looks like it stops at the port name (doesn't seem to allow you to
> [easily] get further WSDL information -- it seems much more appropriate
> to fix binding and or port types names within the WSRP namespace. 
> In the end this means that JAXRPC consumers will [likely] have to manually
> parse the WSDL its beginning to look like they may need to do this anyway
> to support static stubs.  The reason I would recommend standard names
> for both binding and port types is that we are talking about supporting
> factors at both levels.  I.e. the attachments factors is at the bindings
> level while the collection of methods within a given interface is at the
> port type level.  What do folks think?
From the depths of the technical discussion we moved back to the original
question of what types of factors should we support.  I expressed
a desire that from a WSRP perspective we limit the required factors a Portal
would be required to support. I suggested that from the WSRP perspective
we should just require the full blown protocol making simpler representations
optional.  Alan [?] suggested the opposite.  That if we limit
the factors a Portal be required to support that it be the simpler representation. 
We terminated the conference call suggesting that this topic be further
fleshed out via the e-mail discussion list.
    -Mike-