erights
commented
5 years ago For reference, here's the recordings of my two presentations at the summit, with discussion:
Relative Routing https://www.youtube.com/watch?v=a0Wr2BvIixk
ERights Pegging https://www.youtube.com/watch?v=nNImbry1HLM
cwgoes
[ ] Timeouts should close ordered channels
[ ] Implement unordered channels, with causally independent packet timeouts
[ ] Investigate feasibility of DAG ordering on individual packets (including cross-chain) Hunch: more reliable channel ==> efficient acknowledging for multiple messages Unordered channel ==> selective acknowledgement with message hash Packet identities are hash of contents & predecessors, can assume ordering of receipt
[ ] Think about possible fairness guarantees across channels on a connection (maybe for v2) Fairness across set of connections Fairness across set of channels on a connection We can at most restrict validator ordering choices Priority of given message determined by sender
[ ] Clarify IBC relayer module interface, note that it is recommended to users
[ ] Optional acknowledgements of success / failure for packets Module can choose whether or not to register acknowledgement handler Can tell that no reaction will take place, so don't bother acknowledging
[ ] Close "reason" for channels Reason as bytestring for voluntary closure provided by channel System-level reasons for closure (maybe modules can register hooks) Think about cryptoeconomics
[ ] Think about multi-hop case, construct data structures accordingly (e.g. channels are end-to-end) Distinguish between multi-hop, single-hop in data structures, multi-hop not yet supported Maybe trivial "IBC routing" module, maybe it handles loopback
[ ] Review IBC interface for use "slotting in" as VatTP in Agoric's protocol
[ ] Write specification-compliance testsuite with full spec coverage Some level of "binary datagram" -> "binary datagram[s]" How to define testsuite, parse JSON