We have seen a lot of discussions on data availability over the last 2 years, especially in the context of sharding and plasma.
A few of us working on Plasma related data availability solutions thought it would be good to have a gathering during Devcon. Interested attendees could be researchers and engineers working on Sharding and Casper, various Plasma proposals and even state channels. It seems like there are a lot of potential solutions with various tradeoffs that can be reused and it would be great to review and discuss these in person.
If you are interested in attending, please briefly describe the context in which you need to guarantee data availability (e.g. sharding, plasma, state channels, oracles, etc.). This will help organizing the event. In addition, feel free to propose dates, duration, location and event format. This is meant to be an unconference, or “participants-driven” conference.
"Almost there" Schedule Proposal
When: October 28th
Where: HOLIDAY INN Prague Congress Centre, Na Pankráci 1684/15, 140 00 Praha 4, Czechia, Orangerie room
Time : 1h30pm - 6pm
What :
Single, large round table (ideally we are under 30 participants)
Short presentations (~10 mins) by interested parties followed by discussion (20-30 mins)
Presentation should present the system used, why data availability is a problem in this context and the solution being used for now (if any).
Presentations are sorted by topic similarities to optimize discussion flow
I assume that no one who attends this in favor over what is happening at Devcon will need an intro to the problem.
Edit:
Interested in this for all kind of scalibility solutions. Mainly the (plasma) batch auction exchange we are working on at Gnosis (https://github.com/gnosis/dex-research)
I might be able to share some thoughts about a high throughput “proof of availability/ broadcast - only” chain. Similar to what Jason Teutsch is proposing here but instead of POW using bonded validators and also using a DAG instead of a chain.
Agreed. Perhaps just a a round table thing where people introduce in which context they have a data availability issue and how they are currently tackling it? Is a round table with discussions a more appropriate format than presentations?
It is important that the meeting is outside devcon4 building - neither me not any of our engineers who tried were able to buy devcon4 tickets Devcon4 is the worst organized conference I have seen in my life It totally defies my mind why the organizers could not reserve larger space. There are Google and AWS conferences that have hundreds of thousands of developers and Ethereum was able to only host 4,000. I have heard there were 40,000 people that wanted to come and could not.
Sure, would be happy to have this at another location.
I just updated the OP with a list of potential attendees.
For everyone that replied above, could you please briefly describe the context in which you need to guarantee data availability (e.g. sharding, plasma, state channels, oracles, etc.)? Either edit your previous comment or just make a new reply. This would help us organize.
We are building ETH-compatible side chains. In our context, data availability is in the context of block proposals for asynchronous consensus. In our model, less than 1/3 of nodes in a chain are malicious. We need to make sure the block proposal is distributed to the majority of the good nodes before the consensus starts, so if the block proposal is selected to be committed to the chain by the consensus, the majority of the good nodes have it and it can be efficiently pulled by other good nodes. We have a particular broadcast protocol for this, which is mathematically guaranteed to work and uses threshold signatures - I will explain details during the anti-conference)
I’m interested in the data availability problem in the following contexts:
Payment schemes that allow service providers to receive payments once they have generated proofs that some service/computation was executed correctly on some input data. The payer might require the output data, but the service provider can withhold the output data thus creating the data availability problem.
zkSNARK based schemes that involve posting a relatively small amount of data on-chain, but require data to be available in order for users to determine if a state transition they requested was actually processed or not.
I’m working on a state-channels hub-and-spoke model where an operator/hub may generate a “payment forwarded” proof on an offline payment receiver’s channel. This should be about the same problem in @yondonfu first bullet point. I’m looking into potentially solving this with bonded challenges to reveal data as is similar to the plasma cash approach.
I want to join. My team is implementing Data Availability solution for Plasma EVM. But our flight arrives Oct 27th, 4PM. Is it possible to delay a day? or late night would be good.
