I agree with your speaker that it is going to be messy for a while. Big data people are going to be trying to define the system from the top down with verbose and complex protocols, while embedded people are going to lean more towards concise and spartan protocols. Personally, I like the standardized track. Oddly enough, IPv6 has a lot to offer small embedded systems if it is done right. I mark this as odd since the knee-jerk response to IPv6 in parts of embedded is that it is too large to be handled.
It sounds like we are comparing apples and oranges here.
I do think that an IoT protocol battle is happening but there are separate battles at more than one level of the OSI 7-layer framework.
Publish-subscribe protocols (like MQTT, XMPP, DDS, and ARM's Constrained Application Protocol - CoAP) are competing protocol models for data distribution at the Application Layer - pushing sensor data to server nodes. Protocols like MQTT, XMPP, DDS sit on top of a socket interface to TCP/UDP/IP stacks and could co-exist with IPv6 at a lowel level.
Meanwhile IPv6 is competing with other Network layer protocols (like IPv4, and 6LoWPAN - IPv6 for Embedded Devices and Zigbee). 6LoWPAN may be a perfectly valid approach to support a subset of IPv6 functionality on memory-constrained devices.
But I don't think it makes sense to say "why do we need MQTT since we have IPv6" since these protocols are used at different layers with differing responsibilities (apples and oranges).
Can MQTT use IPv6? Absolutely. Client-side MQTT protocol implementations (like mosquitto and RSMB) support IPv6. See the MQTT discussion at http://mqtt.org/2011/09/pubsub-huddle, and the Eclipse IoT site at http://iot.eclipse.org.
Coversant has an XMPP based server which benchmarks at 240,000 concurrent and 60,000 1k messages/second. We will be holding a hands on lab on the EE Live expo floor at 3:00-3:50 on April 1. We will be showing how to develop filters on our server can read message in transit and apply for routing to proper end-point for action to be taken on events.
I hope to have you join us, we are part three of a series that is being hosted by aicas and showing how to get data from device to cloud and controls back.