- From: Martin Gudgin <mgudgin@microsoft.com>
- Date: Tue, 2 Sep 2003 07:29:22 -0700
- To: <xml-dist-app@w3.org>
At the ERCIM face-to-face I took an action item to write up a list of properties we might find a use for in MTOM. Below is such a write-up. Please do not read anything into this mail regarding my personal opinion on whether any, all or none of these properties are actually needed. Also please note that names are just made up for now, we can decide what actual names to use if/when we decide to adopt any of these. Feel free to suggest additional properties I've missed. Gudge Global Properties Sender side: We currently have a single property on the sender side of a message exchange: http://www.w3.org/2003/06/soap/features/abstract-optimization/Optimizati onCandidates which is "A list containing zero or more (pointers to) element information items within the SOAP envelope [which are] candidates for being transmitted in an optimized way." [] is text added by me. Other sender side properties we might consider are: http://www.w3.org/2003/06/soap/features/abstract-optimization/ElementsRe quiringOptimization which would be "A list containing zero or more (pointers to) element information items within the SOAP envelope which MUST be transmitted in an optimized way." This gives control over which EIIs MUST be optimized, while allowing implementations to optimize additional EIIs as they see fit. Downside is that whoever populates this property ( probably the 'application' ) might not know the characteristics of the underlying binding and hence might mandate 'optimization' of EIIs that it actually doesn't make sense to optimize. http://www.w3.org/2003/06/soap/features/abstract-optimization/ElementsTh atMustNotBeOptimized which would be "A list containing zero or more (pointers to) element information items within the SOAP envelope which MUST NOT be transmitted in an optimized way." This gives control over which EIIs MUST NOT be optimized, while allowing implementations to optimize additional EIIs as they see fit. Downside is that whoever populates this property ( probably the 'application' ) might not know the characteristics of the underlying binding and hence might mandate 'optimization' of EIIs that it actually doesn't make sense to optimize. Receiver Side Properties We might also consider a receiver side property: http://www.w3.org/2003/06/soap/features/abstract-optimization/OptimizedE lements which would be "A list containing zero or more (pointers to) element information items within the SOAP envelope which were transmitted in an optimized fashion" Whether this list is exhaustive or a hint is something we would need to decide. Is an implementation required to faithfully fill in this property? Or can it leave it empty? I can see reasons to for either approach. We *might* also want to propogate the sender side http://www.w3.org/2003/06/soap/features/abstract-optimization/ElementsRe quiringOptimization and http://www.w3.org/2003/06/soap/features/abstract-optimization/ElementsTh atMustNotBeOptimized properties over to the receiver. Other global properties http://www.w3.org/2003/06/soap/features/abstract-optimization/Optimizati onType Which would be "A QName indicating the optimization mechanism to be used" currently only xsd:base64Binary is defined. Note we might also need finer grained control over this, see below. This property makes sense on both the sender and receiver sides of the exchange. Other properties The above are all properties that broadly apply to MTOM as an optimization mechanism. We might also want to define properties that apply to specific optimized pieces: http://www.w3.org/2003/06/soap/features/abstract-optimization/EII/Optimi zationType Which would be: "A QName indicating the optimization method used for a given element information item" This is similar to the global property above, but at the EII level. Something like this might be needed if we decide to allow optimization of types other than base64. http://www.w3.org/2003/06/soap/features/abstract-optimization/EII/Conten tType Which would be: "An RFCXXXX media-type specifying a media type for the content of element information item" This information mimics at the property level what is currently present in the SOAP envelope as xmime:MediaType attribute and present in the multipart packaging as a Content-Type header in PASWA.
Received on Tuesday, 2 September 2003 11:05:31 UTC