Host Protocol Stack and Network Service API

Known Issues

• Link layer block id issue on host stack restarts: The Hop link layer data transport uses a monotonously increasing and unique id to identify data blocks. If upstream node (R) has seen a block id before from a host (H, identified by MAC), it simple acknowledges that it has received (and forwarded) the block before and will not forward again. This enables robustness to any misunderstandings whether block was reliably delivered when acks are lost. However, if the protocol stack on host H were to be restarted for some reason it resets block ids to begin from 1 and results in data blocks sent subsequently to node R to be ignored if those ids were used before.
  • Potential fix(es): 1. Use content hash as block id, rather than sequence number. 2.) Modify router to consider repeated sequence number as new if time from prior receive has exceeded a threshold.
  • Workaround: For experiments, restart router nodes when host stack is restarted. Note that all routers need restart and not only first upstream node R since the same Hop protocol is implemented across routers and host stacks.
  • Status: No fix implemented yet.
• Issue X: