[MPC Feedback] No lightweight blockchain data subscription — forced to embed nearcore

[r-near]

Source

MPC team feedback via @marten.blankfors (Slack thread)

Problem Statement

"Lacking a proper solution for subscribing to blockchain data. The fact that the canonical way to do this is to embed a full nearcore node is so painful."

Pain points:

• Ties releases to nearcore releases (every upgrade is a hard fork deadline)
• Hard to debug (nearcore memory leak took down MPC nodes, GC bug filled disks)
• Full node overhead to track a single contract

Research Findings

Root Cause

The MPC team historically chose to embed nearcore directly for lowest latency. This created tight coupling — they depend on 8 internal nearcore crates not meant for external use.

Alternatives Exist

There are several indexing options that don't require embedding nearcore:

• NEAR Lake Framework (S3-based)
• FastNEAR / NEARDATA
• RPC polling via near-api-rs
• WebSocket event streams

Each has different latency/complexity tradeoffs. The MPC team would need to evaluate which fits their requirements.

Remaining Gaps

1. No official push-based subscription — all options require polling (S3, HTTP, or RPC)
2. Sub-second latency — may require running own infrastructure or using optimistic blocks
3. Migration effort — MPC team needs to refactor away from internal nearcore crates
4. Unclear recommendation — no documented guidance on which approach fits which use case

Recommended Actions

• Help MPC team evaluate alternatives for their latency/reliability requirements
• Benchmark latency: embedded nearcore vs alternatives
• Document tradeoffs of different indexing approaches (part of DevEx Indexing initiative)