Explore the overhead of application level transfer protocol
teaguestockwell opened this issue · comments
currently file sharing is implemented by sending an ack for every file chunk, additionally it is using the un ordered protocol of RTCDataChannel. here is what i would like to explore:
- what is the overhead of these acks
- can progress be implemented without acks
- can the RTCDataChannel buffer + waiting for buffer room before sending replace retry
- does the overhead + complexity of sorting out of order chunks outweigh the throughput bonus of unordered RTCDataChannel
commit 4b88997 refactored previous application transfer protocol and reduces memory required to send a file