Purpose of Meeting and important docs 10m:

Important reference documents:

• cid.contact in Kubo
  • https://github.com/ipni/specs/pull/5
  • https://github.com/ChainSafe/go-libp2p-kad-dht/pull/3 (DHT implementation)
• https://cid.contact/
• IPFS Async discussion with leadership on Kubo integration
• Bitswap provider delay optimization
  • https://github.com/ipfs/kubo/issues/8807
  • https://github.com/ipfs/kubo/pull/9530
  • Thunderdome Experiment Summary: provdelayrouting
• Double Hash DHT
  • https://github.com/ipfs/specs/pull/373
  • Chainsafe implementation evaluation: https://hackmd.io/@CUiEDTVxTxaqvvNkKsXWog/rycOOeHdi

Updates from contributing teams 20m:

IPFS 5m

• @lidel team is working on bifrost-gateway binary. Part of Project Rhea. Mentioning here as Saturn backend will dogfood delegated content routing over HTTP .
  • Backend will be filecoin saturn for the time being.

IPNI 5m

• @masih Derkani pushed the DHT cascading solution allowing for lookups across both the IPNI and the DHT without having to run a DHT client to /prod and tested it along with Cloudfront caching! https://github.com/ipni/caskadht/pull/1
• @masih Derkani added refined responses for ndjson to IPNI HTTP API GET /multihash/<multihash> optimized for batch finds https://github.com/ipni/specs/issues/8
• @gammazero submitted a PR enabling saving advertisements and entries from the datastore as CAR files with the ultimate goal of enabling saving advertisement CAR files to S3(PR pending). https://github.com/ipni/storetheindex/pull/1210

Probelab 5m

• Bitswap Provider Delay optimization
  • Thunderdome Experiment Summary: provdelayrouting
    • Outcomes - worst performance with delay optimized to zero, kind of a surprising result! Still analyzing the test methodology possibly a Kubo issue. More to come.
  • Measurements are done, but results are unexpected from a protocol pov (P99 TTFB ProviderSearchDelay=0 is slower than P99 TTFB ProviderSearchDelay=1000)
  • TTFB is strongly correlated with Heap usage.
  • There seems to be weird stuff going on with Heap in use and CPU usage. Maybe something to do with kubo resource manager?
  • Put tests with Bifrost on hold, while we investigate this issue
  • The details on the thunderdome tests are here -
  • Link to details on why - was expected to perform worse -