Please note. There has been a meeting format change intended to make our brief sync time more valuable and more easily understood by new/occasional attendees.
Meeting Lead: @eingenito
Attendees:
- @jimpick
- @hannahhoward
- @aschmahmann
- @michaelavila
- @stebalien
- @momack2
- @gmas
Recording: This recording is private to the PL organization
- Review Announcements for the whole team
- Review blockers and needs from other team members (not general updates)
- Are you blocked waiting for input?
- PRs for review? Remember the project board
- Review progress on major initiatives
- Find or create the section for the IPFS initiative(s) you worked on.
- Summarize the initiative briefly in notes to give a little context.
- Summarize your progress
- Summarize your next steps.
- Review other Questions/Announcements/Notes
@hannahhoward -- out this week Thur + Fri!
Summary: We are investigating various IPFS datastore options with the goal of replacing the current default datastore with a faster, more scalable one: Epic Issue
- This is the GC issue that @magik6k found: dgraph-io/badger#767
Summary: We are improving the performance of transferring data between IPFS instances: Epic Issue
@hannahhoward No Update... but progress starts next week!
Summary: We are improving the API used to drive IPFS functionality and extracting it to a separate project without a dependency on go-ipfs itself. Epic Issue
- merged DI constructor patch
Summary: We are improving the scalability of the providing subsystem (the system for advertising, via the DHT, the content that a given IPFS node can provide). Epic Issue
@michaelavila
- Done:
- Moved existing provider/reprovider behind a common provider system
- Created an experimental flag to toggle between the old and new provider systems
- Came up with plan for small releases of "experimental provider system"
- Now/Next:
- Get the first experimental provider system merged (this first version simply doesn't provide anything and the hope is that its useful on its own for gateway)
- Adding an alternative strategy for providing just roots and get it merged
- Add provider stats command to get some basic information about the provider system (e.g. things being tracked, how much we are able to provide on average, etc.)
- Blockers:
- None, BUT participate in this discussion if you wish: ipfs/kubo#6221
@stebalien
- Done:
- Fixed perf issues in the provider subsystem inside go-libp2p-kad-dht (too many provider subsystems...)
- Actually GC expired provider records.
- Avoid loading provider records into memory unecessarily.
- Concurrent GC.
- Fixed perf issues in the provider subsystem inside go-libp2p-kad-dht (too many provider subsystems...)
Summary: We are converting the default format of content identifiers in IPFS from Base58 to Base32 and encoding additional metadata about the identifer. Epic Issue
@eingenito - no progress this week
Summary: We are improving the use of IPFS via FUSE to provide simple way to consume IPFS content from existing applications. Epic Issue
@eingenito - @djdv was out this morning but we're hoping to see a preliminary PR this week.
Summary: We are analyzing and troubleshooting the performance of the PL IPFS gateway nodes to make then faster.
@eingenito - @gmas to start on IPFS containerization and metrics collection.
Summary: Sync Graphs Quickly (without lots of requests) - ipld/specs#75
@hannahhoward
- Core functionatlity -- DONE!
- Working on docs this week
Summary: We're trying to get better working as a team.
@eingenito:
- Interrupt label to be used to indicate issues that are of high priority and worth deferring long term OKR initiatives to fix/implements. Will review each Monday, and broadcast on Riot.
@momack2
- short 5 min OKR video explaining what we're doing and why
Summary: Make IPNS faster and support more use cases
@aschmahmann
- Continuing to work with @bigs on using the testlab
- Collecting (non-testlab) benchmarks on IPNS-over-Pubsub and trying to understand bottlenecks and how to collect accurate information
- Meeting happened regarding how to make IPNS-over-Pubsub work even faster with a semi-centralized strategy
- Notes: #948
- Follow-up issue on multiple IPNS resolvers: ipfs/specs#198