We have an implementation that implements MPTCP with much much less
intrusiveness than the current implementation. It's review has not
revealed any serious issues. If it's interaction with TCP can be
improved further that is great and patches should be submitted.
I have yet to see any other proposal that would result in any better
implementation and contrary to individual belief's that some folks have,
no one knows what upstream will accept. I bet even David Miller and Eric
do not know what they will accept. That is why they asked for an
implementation to start a discussion, which we have now.
There is no reason to keep working in the dark. Once we have the
discussion there will be ample opportunity to improve the implementation
based on some guideline.