On 3/18/2016 3:27 AM, Josh Howlett wrote: >> E2e places the cost of interoperability on the ends, where it belongs. >> >> The basic principle (it's NOT an argument) is that functions should go where >> they most efficiently belong. I've never seen it that way, but rather: E2E services and capabilities cannot be composed *solely* from their HBH components. I.e., the HBH remains useful for many reasons - efficiency among them. It's always more useful to retransmit over the lossy hop than to wait for the ends to retransmit, but E2E recovery remains required. There's also the problem of defining "end" vs "middle". I use a variant of HW vs SW (i.e. HW = that which you can kick; SW = that which you cannot kick): - endpoint - that which *I* can kick (or SW that runs thereon) - middle - that which *I* cannot kick It hammers home the essence that end/middle is always a relative distinction. Joe