feat: add 2026-06-03 weekly meeting notes (#41)

meetings/2026-06-03_weekly.md

@@ -0,0 +1,94 @@
+# **2026-06-03 MCP: Transport Working Group (Weekly)**
+
+# **TL;DR**
+
+The meeting focused on defining the technical scope of the streaming protocol,
+debating the necessity and scope of session identifiers, and reviewing proposed
+JSON schema constraints. The group aligned on collecting concrete use cases for
+session identifiers on the repository to resolve tension over statelessness, and
+on removing controversial upper limits from the JSON schema hardening proposal.
+Significant technical friction emerged regarding whether to bake logical
+constraints (like non-negative counters) directly into the schema to support
+compiler-level validation and SDK generation.
+
+# **Detailed Summary**
+
+## **Defining the Scope of Streaming**
+
+* **Modality Scope:** Shaun Smith presented a definition of streaming,
+  emphasizing the need to support diverse modalities such as diffusion models,
+  which stream a sequence of progressive frames (denoising noise) rather than
+  text chunks.  
+* **Ephemeral vs Long-Running:** Kurtis Van Gent noted that the streaming
+  protocol must support both ephemeral tool calls (e.g., SQL database queries
+  returning thousands of rows) and long-running operations.  
+* **Tension Point \- Resumability and Transport Limits:** Members raised
+  concerns over how streaming interacts with statefulness. Specifically, if a
+  connection drops mid-stream, the team debated if a resumable mechanism
+  resembling pagination is required to fetch data from the middle of a stream.
+  JSON-RPC over MCP makes handling large binary data transfer particularly
+  challenging.
+
+## **Session Identifiers & State Tracking**
+
+* **Session ID Proposal:** Java Pro (Ripple) proposed introducing a unique ID
+  generated by the client to be carried across requests, enabling server-side
+  state tracking, auditing, and request correlation. Luca Chang noted this helps
+  AWS VM isolation to pin specific containers to sessions.  
+* **Tension Point \- Connection vs Conversation:** Kurtis Van Gent and Stephen
+  Halter argued that a session ID represents a logical "conversation" (which may
+  span different clients or connections) rather than a physical connection.
+  Shaun Smith strongly opposed using the word "session ID," recalling previous
+  hard-won consensus that MCP must remain stateless at the transport layer.  
+* **Tension Point \- Server vs Client Generation:** Stephen Halter argued that
+  the server, not the client, should generate the session ID to prevent cache
+  conflicts.  
+* **Tension Point \- Task Overloading:** Peter Alexander suggested using the
+  existing "tasks" abstraction to track multi-turn workflows. Kurtis Van Gent
+  strongly opposed, arguing that tasks are finite "promises" for single results
+  and overloading them adds unnecessary complexity.
+
+## **JSON Schema Hardening & Logical Constraints**
+
+* **Schema Hardening Proposal:** Andreas Schlapbach proposed adding strict
+  constraints to the JSON schema, such as maximum string lengths and ensuring
+  counters cannot be negative, to mitigate DoS attacks and optimize SDK code
+  generation.  
+* **Tension Point \- Opposition to Size Limits:** Stephen Halter and others
+  strongly opposed hardcoded upper bounds (such as a 64k state limit), arguing
+  that arbitrary limits become obsolete and that size defense is an
+  implementation concern for individual servers, not the MCP spec.  
+* **Tension Point \- Logical Schema Constraints:** Andreas Schlapbach shifted
+  the focus to logical constraints (e.g., ensuring a progress index is
+  non-negative, or array lengths cannot be less than zero). Kurtis Van Gent
+  questioned the need to specify common-sense limits, but Andreas Schlapbach and
+  Peter Alexander clarified that defining these in the official JSON schema
+  allows compilers to automatically generate type-safe validation annotations
+  and code in languages like Java and Go.
+
+# **Next Steps & Timeline**
+
+Kurtis Van Gent will cancel the next weekly meeting due to travel. Shaun Smith
+will revise the streaming problem statement to include concrete database query
+use cases. Java Pro will open a dedicated tracking issue on the Transport
+Working Group repository to gather documented use cases for request correlation
+and conversation tracking.
+
+# **Action Items**
+
+- [ ] Shaun Smith Revise the streaming problem statement to include concrete
+  database query examples and progressive frame use cases.  
+- [ ] Kurtis Van Gent Open a tracking issue on the Transport Working Group
+  repository to collect concrete, non-abstract use cases for
+  session/conversation identifiers.  
+- [ ] All Participants Post specific, documented use cases of session identifier
+  needs to the newly created tracking issue.  
+- [ ] Andreas Schlapbach Contact Shaun Smith and Ola on Discord to review and
+  align with existing JSON schema tightening efforts.  
+- [ ] SDK Teams Review the proposed JSON schema updates to ensure they do not
+  negatively impact SDK compiler generation.  
+- [ ] Andreas Schlapbach Update the protocol pull request to remove the
+  controversial upper-limit size constraints.
+
+Source: [Meeting Notes — June 3,
+2026](https://docs.google.com/document/d/12Z_Eg7sFXBDGuQaiQWSNi3vRTlc3XI_T3HoqLli0HoA/edit?usp=drive_web&tab=t.4ww5di3nev3f)