RE: More thoughts on contexts (was RE: [ws-caf] Agenda for the demo application)

["Jim Webber" <Jim.Webber@newcastle.ac.uk>]

Hi Alastair,
 

> Timeouts: it all depends what the timeout means. The hard 
> part is the interpretation, and I think that must be known to 
> the specific processor, not to the general contexts 
> processor. So once again, we seem to end up employing a 
> servant to open the door, when we are perfectly able-bodied.

The interpretation of a timeout at the WS-Context level is easy enough.
After the timeout period, the context is no longer valid. What that
means to an application is application-specific.
 
> A standard model for hierarchical contexts is more 
> interesting. One question I have is: how easily can I 
> construct disjoint hierarchies?

Pretty easily enough I would say. 

> Let's say that the interposition concept is useful for 
> transactions, but we want a flat security domain. I'm not 
> sure what that would mean in practical processing terms 
> either (carried together in a contexts wrapper? Separately?).

Separate context blocks since they are separate contexts for separate
domains.

> Related question: in WS-TXM there is heavy emphasis on 
> interposition, which is an example of hierarchical context. 

I didn't think the emphasis was very heavy, but it's in there because
transactions people reckon it's useful.

> However, I cannot discern from the specs how the WS-CTX 
> hierarchy is used to help this.
> Specifically, the WS-TXM protocols appear to defer to WS-CF 
> for interpositon, but WS-CF does not seem to mandate any 
> defined relationship between a Coordination Service and an 
> ALS -- which seems to be the point of contact with WS-CTX as 
> it currently stands.

I'm not sure I see the problem here. WS-Context provides the base
context for the stack. It supports hierarchical contexts well enough.
WS-CF contexts are an extension of WS-Context contexts, and so they too
support hierarchies. Same again for WS-TXM contexts.

WS-TXM sits on top of WS-CF which sits atop WS-Context. Given that ALSs
are the point at which lifetime events for contexts and that
coordination happens at specified points in the context lifecycle, I
don't see any confusion there either.

However, I have to add that in my current work I am only interested in
WS-Context, and so others who are working further up the stack (like
Arjuna) might shed more light on this.

Jim