Message parsing   

I guess you are referring to this topic: http://publib.boulder.ibm.com/infocenter/wmbhelp/v6r1m0/topic/com.ibm.etools.mft.doc/ac18870_.htm

1. If you select a Parse Timing value of Immediate, on-demand parsing is overridden, and everything in the message is parsed and validated except, if the message domain is MRM, those complex types with a Composition of Choice or Message that cannot be resolved at the time
2. If you select a Parse Timing value of Complete, on-demand parsing is overridden, and everything is parsed and validated. If the message domain is MRM, complex types with a Composition of Choice or Message that cannot be resolved at the time cause a validation failure.

I think you have misunderstood what this is saying. The MRM parser can resolve and validate complex types. It can even resolve choices automatically and resolve complex types with Composition="message". It cannot, however, resolve a choice which does not have a tag. The validation options are provided for those rare situations when the message flow needs to resolve the choice, but the user wants validation of the other fields in the message.

If the CWF parser encounters a choice, it
a) calculates the maximum length of the choice ( the length of the longest member of the choice ). Call this 'maxChoiceLength'
b) extracts 'maxChoiceLength' bytes from the input message, as a BLOB
c) creates a NameValue element in the message tree and sets its value to the extracted BLOB.

The item created in c) is an 'unresolved choice'. The ESQL/Java/PHP code can resolve that choice by referencing one of its branches. This provides the parser with the missing piece of information ( i.e. which branch it should parse ). It then replaces the NameValue element with the elements of the resolved choice branch.

That should answer your questions about the meaning of these validation options.