Our terminology got somewhat distorted. In your SOAP example, soapaction has the same role as port-ID (subject-ID or service-ID) in UAVCAN. There is no need to identify the data type explicitly because the data is represented using a self-describing encoding (XML). I don’t think SOAP is hugely relevant though — my references to SOA are related to the design principles rather than one particular method of implementing them in a practical system.
I did spend some time pondering about your latest posts but I am not yet ready to offer much valuable input. I do have a question though.
Speaking very generally, if the type information was trivially deducible from any received transfer, would you find such design acceptable? Your proposal of adding semantic-ID to DSDL suggests that it might be the case.
I should leave this thread here which contains my unedited thought process behind the current architecture:
""
who knew that happiness can be measured in bytes