Hi Noel, Thank you for highlighting this document. I had seen it before and the implementation (as found on the tapes from CSRG and now on THUS) follows the plan outlined in IEN168 quite closely. The first snapshot of the code is just a few months after this document. In a way it is modeled after the UoI Arpanet Unix implementation (and thank you again for finding that source!), with a separate (kernel) process for network activity. In my experiments I have found that it is not all that easy to get smooth network data flow as this network process is difficult to schedule just right. I now better understand why Joy moved to "software interrupts” to get better scheduling of kernel network operations. Wbr, Paul