Roadmap meeting notes:
Attendees:
- Medha Parlikar
- Rita Allen (Unlicensed)
- Former user (Deleted)
- Mateusz Gorski
- Former user (Deleted)
- Tom Vasile (Deactivated)
Goals:
Review the proposed roadmap milestones & release dates & high level features: https://docs.google.com/spreadsheets/d/1SwHALHiP03Ocglom-HB3RlzZROHVYmhuLi6pECFVUBY/edit?usp=sharing
Notes:
Notes from meeting | Question | Response |
---|---|---|
Hiring is a concern - given the current team size, the roadmap is probably not achievable. | Based on the roadmap, we need to focus our hiring on:
| Rita - ideas:
|
Consensus & testing represents a huge chunk of work
| What features were not complete with RChain PoS? What does "not correct" mean? What do we want to do about it? | Medha Parlikar |
Change Data Retrieval to → Client API's | Can we say "node API"? | |
Need a row for Formal Verification in CON - (from meeting notes)
| This row created a bit of confusion - what exactly are we verifying and who do we expect will do this work? | |
Row in EE for formal verification (from meeting notes)
| Again this row created a bit of confusion - who do we expect will do this work? | |
What is the vision for DevNet? We made this assumption: DevNet - supports demonstrable features from previous month, supports performance testing. Given this, we removed the concepts of "internal" and "external" devnets. | ||
Virtual DNS/hiding IP addresses of validators - is this our responsibility or the app owner? |
Changes to Roadmap:
- Move Contracts from Jan to Feb/March
- hard code validator set for January & Token Contract for Jan.
- What is 'createblockstatus'?
- Comes from Mateusz's node architecture
- Updated Roadmap
- Change Data Retrieval to → Client API's
- Pull in PR's from Daniyar in RChain repository
- Rita Allen (Unlicensed) - Tasks on tap for Abner to pull in his PR's
- Metrics
- Need a single page of requirements for Metrics.
- Former user (Deleted) - stepping up to clean up the metrics pages - Add a section to the specification. /wiki/spaces/AR/pages/3014657
- Deploy gossiping
- Comms specification
- Consensus - how do we deal with overlapping deploys? Need a specification
- Added a column for stabilization, planning for likely learning that will happen when the features comes together.
VM will not care - executing contracts (token or not token);
Blessed contracts - PoS, token,
Consensus - requires a set of validators, can be static or dynamic. Start with static, move to dynamic. Tokens need to exist, but may not be exchanged.
Parallelism: two levels, execution and DAG. Can execute serially but build a DAG; Can execute in parallel but build a linear chain.
Executes aggregates - user pays for execution (and it is persistent)
bonding - validators can join network
From Jack Mills:
axioms (let me know if you agree):
- we are more likely to accurately predict the progression of feature sets than the timing of said progression.
- there is an inverse correlation between level of detail and accuracy, i.e. the more detailed the projections the lower the accuracy.
- there is a direct correlation between time and accuracy, i.e. the further out in time the lower the accuracy.
therefore I suggest:
- first create a time-independent progression of macro feature sets (“macro” = treating the node mostly as a black box as seen by operators and dapp developers)
- then use this to drive lower level feature progression
- then map to time: do monthly increments for the first N months, then quarterly for the next M quarters, then (if necessary) feature set progression without time
my swag at macro node functions:
- contracts
- accept and execute new contracts from clients
- validate and execute contracts from other nodes
- query dag state
- distinguish between different types/modes of contract, e.g. only changes its own persistent state vs. sends messages to other contracts?
- other phases of contract lifecycle?
- wallets
- ?
- distributed consensus
- only applies to multi-node
- storage
- operational
- single node
- multi-node, no node failures, well-behaved peers
- multi-node, node failures, well-behaved peers
- multi-node, node failures, fallible and/or malicious peers
are there any implications of immediate fundraising on the roadmap?