context

Decentralized Service Engineering

this is a point-to-point protocol for:

• large data queries
• high level service interactions like
  • nodes in a cluster,
  • PoP’s in a network,
  • our own cross-data center traffic,
  • bucket-vm

This protocol captures the requirements we face when optimizing workloads to run

• in multiple data centers
• with a high degree of cachability
• to download directly from ObjectStores when appropriate
  • like S3
  • and R2

According to Adin, we have the space to start experimenting with new transport protocols that can be discovered via cid.contact using features available to us in store-the-index.

We can start experimenting today with protocols that are highly optimized for our constraints and cost structures, with a solid path to defining it as a widely accepted transport protocol for large providers like us after we’re happy with it.

We can even release this API in alpha to our own users in need of faster reads of large data, or more advanced queries like we have planned for w3query.

fast-ptp

This protocol uses HTTP verbs and cache control semantics but could be implemented in other transports.

This is not another “Gateway” protocol. It’s optimized for large reads and new query protocols implemented as UCAN invocations.

GET /{multihash}

It is expected that a provider will return a 404 if the data is not easily available in that provider.