search

omgnetwork / research

43 stars 2 forks source link

Reduce Minimum Finalization Period #96

Open boolafish opened 5 years ago

boolafish commented 5 years ago

As prioritized in Cinnabon priorities

Description (From the priority doc)

Current exit period is 7 days, this means a finalization outside our plasma network would need such period. What can we reasonably reduce this time to? What are the implications if the exit period were to be reduced to 24 hrs.

PS. The reason I’ve picked 24 hrs is that in most traditional markets settlement T+1 is acceptable (and faster than what happens for OTC most of the time).

Business Indication

(from slack) Technically speaking, the longer the exit period -> the better guarantees (safety). Taking business point of view, the longer period -> the more time funds cannot be used. In other words, it's better from business perspective to have a short period. Maybe we can follow-up on this in regards to "fast withdrawals" from the Research Priority file next week or whenever we have some free time? (edited)

Why Finalization in Plasma is not enough?

boolafish commented 5 years ago

As I commented in the doc, the biggest issue of naively reducing the exit period is how many utxos we can have in the whole network to support mass exit. Reduce from 7 to 1 basically means we can support 1/7 of our max utxos that is secure currently (which is around level of million currently)

paulperegud commented 5 years ago

Fast withdrawals with help of liquidity provider seems to be a nicer way to handle the problem. It just needs to be fully automated in client software.

boolafish commented 5 years ago

btw, I can think of an interesting research topic is that can we have different exit period on different tx type? And what does this implies?

But I do agree fast withdrawal is our current best bet on this.

paulperegud commented 5 years ago

btw, I can think of an interesting research topic is that can we have different exit period on different tx type? And what does this implies?

I don't see this as working.

  1. Operator can create sybils.
  2. Sybils can assume whatever role that is allowed to create shorter exit period transactions.
  3. Operator creates and withholds block from which she exits using privileged transaction.

There is effectively only one exit period and it is there to address block withholding problem.

paulperegud commented 5 years ago

Reduce from 7 to 1 basically means we can support 1/7 of our max utxos that is secure currently

We should also revisit mass exits to alleviate this problem.

boolafish commented 5 years ago

Related plasma build discussion: https://plasma.build/t/plasma-withdrawal-time-is-a-major-hindrance-to-adoption/66

tousthilagavathy commented 5 years ago

I'm the Thilagavathy that posted the "Plasma withdrawal time is a major hindrance to adoption" and proposed solutions, as in the link below.

https://plasma.build/t/plasma-withdrawal-time-is-a-major-hindrance-to-adoption/66

I like what OMG is working on and have followed it well. It is what led me to concerns and details about Plasma adoption. I'm also part of ONC (OMG Network Community) and in the ODP.

Reiterating my points

. Plasma withdrawal time of days/weeks especially with the DEX use case is a big turn off for users and hence a major hindrance to adoption. Centralized exchange withdrawals happen in seconds/minutes. So, I strongly feel users will not wait for days/weeks for a withdrawal to happen.

. Fast withdrawals with a liquidity provider as discussed in the post also appear not to be a practical solution.

Solutions are discussed in the post. What is your opinion. Do you think it will help to solve the problem?

boolafish commented 5 years ago

Hi, @tousthilagavathy great thank for the post and the points!

I think we are seeing ideas on centralized liquidity provider at this moment: #99. I personally agree with you a liquidity provider market might not work that well, especially for small coins.

Leveraging SGX might be another option for us at this moment: https://github.com/omisego/research/issues/95#issuecomment-498098122

As for your ideas, I will hold my comment on the POS one at this moment cause it is unclear and need more time for us to know how do we want our POS plasma to look like.

For this one:

If watch towers were provided sufficient incentives to be continuously online, they can challenge the fraud immediately and can the withdrawal time then be reduced to an hour or lesser?

It seems to me it is possible in plasma cash(flow). However, in M(ore)Vp we cannot do that because it limits the total utxo throughput. see the discussion here above.

tousthilagavathy commented 5 years ago

@boolafish thanks for the reply. Hope you guys find a good and practical solution to the problem.