Creator
Created
Aug 31, 2023 2:01 PM
Problem:
- A practical fast retrieval PoRep must be in the latency model, where the deadline to submit a proof is as small as possible (ideally <1s).
- The smaller the window to submit a proof, the faster the PoRep initialization step can be.
- To achieve <1s PoRep reply time, either:
- Execution extension: use VDFs to repeate multiple smaller PoReps in a non-interactive way
- or, Faster block time: we must have a faster block time (an fast transaction inclusion)
Problem: can we have fast timestamping of proofs?
Solution 1: Filecoin mainnet with sharded power table to support fast block time
- Why do we need regions? To lower down block time as small as possible
- Open problems:
- Will a sharded power table be secure?
- TODO: Show that power table will not be secure
- Will we have 300ms block time?
- TODO: Write a short paragraph and give me the link to the data
- Showing that 300ms for a region of nation-wide can’t work
- Suggested realistic latency
- With a X%
- Can we guarantee that a tx gets in in next block?
Solution 2: An IPC (or parallel chain) to timestamp
- Ways to get more realistic latency:
- The more the regions the better the latency
- Requirements
- must be 300ms blocktime 90% of the time
- it should support at least 100 nodes
- distributed either across the globe (or grouped by continents)
Mainnet
Randomness
Aggregator network
Timestamp chain
Storage market subnet