feat: add 2026-01-14 weekly meeting notes (#4)
Add 2026-01-14 weekly meeting notes.
This commit is contained in:
parent
dabcc607a5
commit
2ef724d28c
1 changed files with 75 additions and 0 deletions
75
meetings/2026-01-14_weekly.md
Normal file
75
meetings/2026-01-14_weekly.md
Normal file
|
|
@ -0,0 +1,75 @@
|
||||||
|
# 2026-01-14 Transports WG (Weekly)
|
||||||
|
|
||||||
|
## TL;DR
|
||||||
|
The group reviewed positive progress on SDK alignment, confirming a
|
||||||
|
cross-language agreement on bindings. A GitHub repository has been established
|
||||||
|
to document outcomes and manage weekly notes via Pull Requests. The group
|
||||||
|
outlined a timeline targeting a final proposal vote by the end of March.
|
||||||
|
Significant discussion focused on the HTTP standardization proposal,
|
||||||
|
specifically the use of standard and custom headers to improve routing
|
||||||
|
compatibility with existing infrastructure like load balancers. Debates centered
|
||||||
|
on security implications, header limits, and the trade-offs between client-side
|
||||||
|
complexity and server-side scalability. Microsoft offered to serve as a test
|
||||||
|
party for the multi-roundtrip request track to address current pain points with
|
||||||
|
SSE and elicitation.
|
||||||
|
|
||||||
|
## Detailed Summary
|
||||||
|
|
||||||
|
### Roadmap and Process
|
||||||
|
* **SDK Alignment:** Kryspin Ziemski reported a positive connection with Felix,
|
||||||
|
confirming alignment on cross-SDK and cross-language bindings. The team will
|
||||||
|
collaborate on uniform documentation.
|
||||||
|
* **Documentation:** Peter established a GitHub repository containing
|
||||||
|
folders for docs and proposals. The group agreed to use this repo for
|
||||||
|
documenting outcomes and managing meeting notes via PRs to allow for community
|
||||||
|
review.
|
||||||
|
* **Timeline:** A project timeline was outlined:
|
||||||
|
* End of January: Initial proposal to core contributors.
|
||||||
|
* End of February: Reference implementation to validate the proposal in at
|
||||||
|
least one language.
|
||||||
|
* End of March: Finalize proposal for an official vote and acceptance into
|
||||||
|
the specification.
|
||||||
|
* **Track Management:** A table was created to check in on the status of various
|
||||||
|
tracks. The TypeScript SDK was designated as the "proving ground" for
|
||||||
|
prototyping and design work.
|
||||||
|
|
||||||
|
### Technical Decisions: HTTP Standardization
|
||||||
|
* **Header Proposal:** The group discussed a "straw man" proposal to standardize
|
||||||
|
HTTP headers (e.g., `MCP-Method`, `MCP-Tool-Name`) to expose critical routing
|
||||||
|
information. The goal is to allow networking proxies and load balancers to
|
||||||
|
route requests without parsing the entire JSON RPC payload.
|
||||||
|
* **Custom Headers:** A mechanism was proposed to mirror specific tool
|
||||||
|
parameters into sanitized headers (`X-MCP-Param-Name`) to support routing
|
||||||
|
logic, such as regional database requests.
|
||||||
|
* **Security Concerns:** Participants debated potential risks, including "poison
|
||||||
|
pill" attacks where malicious servers could demand an infinite number of
|
||||||
|
headers to exhaust client resources. There was strong advocacy for
|
||||||
|
comprehensive conformance tests and potentially establishing reasonable limits
|
||||||
|
on required headers.
|
||||||
|
* **Routing Alternatives:** The group compared header-based routing against
|
||||||
|
path-based routing and query parameters. While path-based routing matches some
|
||||||
|
existing framework patterns (e.g., Flask), headers were preferred for avoiding
|
||||||
|
payload "cracking" at load balancers. Query parameters were dismissed due to
|
||||||
|
lack of support in gRPC.
|
||||||
|
* **Schema Definition:** Discussion occurred regarding where header extraction
|
||||||
|
metadata should live. While the tool description schema is the natural place
|
||||||
|
for parameter properties, some argued for a separate section to keep the
|
||||||
|
schema focused solely on data definition.
|
||||||
|
|
||||||
|
### Multi-Roundtrip Request Track
|
||||||
|
* **Stateless Protocol:** Clarification was provided that discovery cards fall
|
||||||
|
under the "stateless protocol" track, while changes to elicitations and
|
||||||
|
sampling belong to the "multi-roundtrip request" track.
|
||||||
|
* **Synergy with Tasks:** It was noted that the tasks workstream overlaps with
|
||||||
|
multi-roundtrip requests, offering an opportunity to simplify task workflows
|
||||||
|
using a stateless approach.
|
||||||
|
* **Testing:** Caitie McCaffrey offered Microsoft as a test party for this track
|
||||||
|
due to real-world challenges with Server-Sent Events (SSE) and elicitation.
|
||||||
|
|
||||||
|
## Action Items
|
||||||
|
- [ ] All track leads/participants to write a brief status summary in the track
|
||||||
|
check-in table.
|
||||||
|
- [ ] Kurtis Van Gent to consolidate feedback on the HTTP standardization
|
||||||
|
proposal and work with Mike to upload a version to GitHub for further comment.
|
||||||
|
- [ ] Mark Roth to prepare a presentation for the "multi-roundtrip request"
|
||||||
|
track for next week's meeting.
|
||||||
Loading…
Reference in a new issue