Hey all, as I mentioned in my last post I haven't been able to make the past few hangouts. There wasn't a transcript of the October 2nd one on Beyond Bitcoin but I also didn't have copious amounts of time today; so here's a summary of that one. []'s indicate editor notes and keep in mind this is all my interpretation I could've got something wrong (but I don't think I did.)
(Code tags were the only way to make the tabpsaces big enough for this to be readable. Copy it to notepad or something for an even better experience.)
- Thanks for testing corner cases.
- Testnet discovered blackswan
- Killed network [hardfork required]
- Wasn't supposed to be able to happen
- Every new block failed for the same reason
- Now fixed
- Accounts can vote for vary numbers of witnesses
- Default behavior was to vote for the minimum number
-Lazy voter turnout (like now) would result in minimum number of witnesses
- Now, if you don't vote for at least 2 people you defer to those who do so for the number of witnesses that you will vote for
- If you wish to abstain you can abstain, whereas before you couldn't
- New, fixed testnet is already up
- Very successful testing
- Network stayed in sync w/ floods
- Margins calls, order matching, forced settling (aside from black swan) all worked well
- UI has continued to improve
- Permissions added
- Add/remove multisig keys
- Memokey
- Creating/Issuing assets fixed
- Help integrated into wallet
- Content is sparse at the moment
- Confident for Oct 13th launch
- Deploying infrastructure
- Making sure exchanges are prepared
- Xeroc has some docs for them
- They have been following the upgrade
- Statement will be requested
- Safest thing to do is hold funds locally
- Releasing soon:
- Hosted wallet, testnet
- Light-weight wallet, downloadable executable
- Full-node wallet, for mac, executable
- For windows will be DIY
- Built version coming ASAP
- Light-node or full-node offer best security
- Working on getting referral program in place
- Only very minor GUI changes from now to launch
- Wallet is looking good working under Graphene
- After Oct 13th still providing updates on a weekly basis
- New GUI release first
- No hardforks after Oct 13th expected
- Many things to be added [requiring hardforks], going to be judicious in doing so is a big disruption to the entire network
- Hardfork to add features once a quarter, unless fatal bug
- Metaexchange, blocktrades (Integrate bridges into the web wallet?)
- Working with blocktrades
- CCEDK
- Close to release, is how to connect to network going to change?
- Livenet will be similar to test
- Need to checkout from BitShares repository (not testnet repo)
- BitShares 1 can stay alive, but Bytemaster sees no point in doing so
- Have BitShares 2.0 witnesses been selected?
- Start with initial witnesses voted in
- Publish seed node
- Start election
- Were unable to migrate votes from BitShares 1
- Thus many default witnesses at first
- They do not plan on running witnesses for the long-term, just the transition
- Why can't delegate votes be moved over?
- Fundamental architectural difference
- Every balance not tied to account name, has balance in 1.0
- Every balance tied to account and has balance in 2.0
- Also 1 has many stale votes, new election would be good
- The risk is high with CNX being the initial witnesses
- We will select people we know [trustworthy people] to be the initial witnesses
- Not necessarily CNX employees
- Connect to backbone nodes exclusively. DDOS protection.
- Hasn't been implemented, would like to eventually
- Long-range nothing-at-stake attack against initial witnesses
- Public record out there makes it so that you won't trust any other chain, regardless of length, everyone knows what the real chain is
- Vitalik, weak subjectivity
- TaPoS
- Not a worry for any Graphene DPOS chains
- Long-range nothing-at-stake: someone who used to have control offer block production can create an alternative chain and isolate a victim and make them think they have been sent real money
- User's account wouldn't exist on their chain
- If it did, your keys wouldn't be the right keys
- See Bytemaster's blog post on long-range nothing at stake
- Based on how rapidly witnesses can be voted in on testnets, 30+ in 24 hours
- Once checkpoint made, (~24 hours), long-range attacks can't be executed
- Checkpoint: universally known good state
- Full-nodes have these
- Eventually community will be able to hire competing actors?
- Bytemaster: Yes.
- Cryptonomix is producing the software (free speech)
- Other people run nodes
- Spreads legal risk by separating these tasks
- Software is open-source, so no trust is needed
- Best regulatory protection by not needing trustworthy
- Worker proposals: When worker's ask for shares/share dilution, are they setting a date in the future on which their shares will mature? Or a 100 day sliding window? [Does payout come all at once or over some time-frame]?
- Sliding window.
- Pros/cons of allowing workers to bundle vested shares into savings bonds instruments that could be traded to raise capital to fund something; like a sovereign bond, network itself is guarantor. Market will look at trust of whole network and set price for such bonds.
- Bytemaster: Almost possible by transferring account of worker to cash provider and cash provider gives shares immediately
- Like OTC transaction
- Still on network, but finding buyer
- Similar to bond sale (country, corp)
- Bytemaster: Right
- [Being able to trade vested shares from one user to another.] No bid/ask market? Because bonds are bundled together with shares that have different maturity dates?
- Bytemaster: Vesting system is more robust than that
- Not a bunch of different bonds, one for each day
- Each day you're adding to the bond
- Can withdraw after accumulating
- For each coin-year can withdraw 1 coin
- 365 coins, each day can withdraw 1
- Similar to vested shares from merger, adding another restriction, incentivized to wait out the whole period
- Bytemaster: If you keep it all there you get to all fastest, if you take out half takes twice as long to get the second bit
- Shareholders could appoint trustees or workers who bid on this, raise funds based on the trust in the network and sell promissory notes (dilution in the future)
- Network can borrow in the present and not feel the effects of dilution until the future
- Bytemaster: Very complex economic scenario
- Price in present based on sell-pressure now
- If someone plans to hold for 1 year and they swap for a bond, so someone else can sell in the present
- There's still selling in the present
- Just added a bit of guaranteed long-term holding
- Discounted with time-value money, other factors
- How is the sell pressure of a vested share different from the sell pressure of an actual present, existing share?
- Bytemaster: Accounts can be transferred, vesting account balances cannot be directly transferred
- It would be very easy to add this functionality
- Vesting funds aren't fungible though, and there's no market for it
- Could be done off-network
- Mumble / live-stream on Oct 13th, 2015?
- Fuzzy: Will look into it
- Connect Google hangouts, mumble, Skype
- Bytemaster: On 13th we are doing open-heart surgery
- A lot of infrastructure role-out
- Differentiate day of starting network from grand opening
- Issues may occur
- Have grand opening once everything has been working well for a while
- 13th is like beta release
- UI still needs work
- Bugs will be discovered
- Fuzzy: Now we're going to open-beta, anyone can be a tester
- Bytemaster: Emphasize beta
- There're so many unknown-unknowns
- Only real test is length of time in market
- This is why Bitcoin doesn't change anything
- Few "emergency days" would not be unsurprising [hardforks?]
- More unit-tests than ever before
- Every feature tested at protocol level
- So many variables
- Even in case of hardfork, funds are safe if private keys are safe
- Exchanges will use delayed node that looks back several blocks
- Blackswan problems tend to stop the chain
- Still confident in system
- Audience: Team has good track-record with dealing with such issues
- Fuzzy: Bitcoin is static because they are stable. When will BitShares 2.0 achieve this?
- Bytemaster: Individual call.
- Once every op has been used once, every scenario has happened would inspire confidence for me
- Others will need more
- 6-months seems like a good time for vetting existing features
- Hardforks may reset this clock if they change these features
- We're likely to see issues in month 1 or not see them at all (electronics failure curve)
- Audience: Simultaneous dev chain? Run features on dev chain for 3-6 months before consideration of forking into main chain?
- Looking at ways of making the blockchain technology at the core robust and unchanging and allowing features to be added without breaking things from a blockchain perspective [separation of concerns].
- The challenge with consensus is that anything that changes ownership of tokens changes everything after it. Butterfly effect.
- Bytemaster: Some smart people have been criticizing our claims of 100,000 transactions per second.
- They are attacking a strawman
- Addressed this on the forum
- "We keep everything in RAM and at 100,00 TPS that's a terabyte of RAM per day that would need to be added - this doesn't scale."
- This would not work.
- Not all transactions are kept in RAM though
- Only the state, account balances
- 99% of transactions are simple transfers and result in no net increase in the [complexity of the state or] RAM usage
- I analyzed this from a perspective of, if each account has 1 KiB of data, which is a bit
- Their pubkey, balances, assets, open orders
- Most accounts use much less
- Used 1 KiB as average
- All account information for 1 billion accounts in 1 TiB of RAM
- Can support 2 TiB today
- By the time we grow to be this big, RAM will not be a barrier we will face
- Sequential processing bottleneck sets transaction rate limit (market operations, transactions that impact the market)
- Everything else can be done in parallel
- No way of getting around this within the market
- Every op affects the order book, can't do things in parallel without parallel markets
- Spreads, more complexity
- Parallel chains, one consensus set: side-chains approach. Only accelerates some tasks (cores vs speed CPU analogy - cores only help with parallel tasks)
- Probably max out at a couple thousand transactions per second right now (whole system perspective)
- Have seen couple hundreds of TPS in testnet
- Far in excess of what we're likely to generate in the near future
- If we did we'd be deflationary immediately (more funding, could get needed infrastructure)
- Marketing should be refined to reflect this TPS business
- Minimum requirements for witness VPS?
- Bytemaster: Digital Ocean 1 GiB node has worked for some
- No more than couple hundred megabytes of RAM
- Almost any computer
- Bytemaster summary: Working on deployment issues
- CCEDK
- Exchanges
- Forking codebase
- Setting up for release
- Please test on testnet this week
- Test wallet soon
- Several iterations, dry run of open-ledger wallet
- Have reached out to exchanges
- Get others in contact with Bytemaster
- Xeroc also working on exchange integration
Topic: BitShares Dev Hangout: Bytemaster – October 2, 2015 Detailed Summary (Read 1844 times)
