RE: FW: Reliable Messaging Summary

see below.

> -----Original Message-----
> From: [email protected] 
> [mailto:[email protected]] On Behalf Of Geoff Arnold
> Sent: Wednesday, March 05, 2003 6:22 AM
> To: [email protected]
> Subject: Re: FW: Reliable Messaging Summary
> 
> 
> 
> Just a couple of parenthetical comments about reliable messaging.
> 
> (1) "Reliability" is not a single goal, an objective function 
> that we are seeking to optimize. It describes a spectrum of 
> possibilities. If we accept that  there are costs associated with 
> different
> levels of reliability, we must accept that there may be cases 
> where application deliberately choose a lower reliability to 
> satisfy their overall requirements.
> 
> (2) The requirement of transport independence (including
> the dynamic composition of multiple transports) means that
> at the web services architectural level the specification, 
> description, and negotiation of reliability cannot be couched 
> in transport-specific or even message-specific terms. (If my 
> messages are conveyed using a black-box service that 
> guarantees the required reliability, I should not have to 
> understand exactly how it achieves
> this.)
> 
> If we put these together, it seems that a discussion of 
> reliability at the web services architecture level should 
> involve the development of an ontology of reliability as well 
> as standard mechanisms for describing and negotiating various 
> levels of reliability. These might be expressed in terms of 
> SOAP and WSDL constructs and negotiation MEPs.

agreed but the mechanism for describing and negotiating should not be 
specific to reliability. Other qualtities can use the same framework,
and dare I suggest a ws-policy like
framework needs to be part of our architecture.

> 
>