feat: add 2026-04-29 weekly meeting notes (#37)
This commit is contained in:
parent
14434c45a5
commit
fd7e26ab79
1 changed files with 80 additions and 0 deletions
80
meetings/2026-04-29_weekly.md
Normal file
80
meetings/2026-04-29_weekly.md
Normal file
|
|
@ -0,0 +1,80 @@
|
|||
# **2026-04-29 MCP: Transport Working Group (Weekly)**
|
||||
|
||||
# **TL;DR**
|
||||
|
||||
The meeting focused on the stateless transition timeline, SDK conformance
|
||||
testing versioning, and proposed feature agendas for future releases. Discussion
|
||||
centered on the transition of the initialization flow to `serverdiscover` under
|
||||
standard IO and the introduction of an E-tag mechanism for optimistic
|
||||
concurrency correctness. Tensions arose regarding whether standard IO can
|
||||
natively support headers for features like E-tags, leading the group to identify
|
||||
the need for exploring HTTP over standard IO. Proposals for streaming and raw
|
||||
binary handling were identified as priorities for the release after next, while
|
||||
encryption and compression proposals were tabled due to a lack of immediate
|
||||
customer demand.
|
||||
|
||||
# **Detailed Summary**
|
||||
|
||||
## **Stateless Migration & Conformance Testing**
|
||||
|
||||
* **Initialize Flow Removal:** The transition to a stateless architecture
|
||||
involves removing the initialization call in favor of a `serverdiscover` (or
|
||||
`initialize server discover`) call to determine protocol versions and remove
|
||||
session constraints.
|
||||
* **Testing and Merging Timeline:** The final lock-in date for SEPs is April
|
||||
30th, and approved SEPs must have conformance tests merged by May 15th before
|
||||
they are finalized at the end of June. All Tier 1 SDKs (TypeScript and Python)
|
||||
are expected to implement changes by the release date.
|
||||
* **Protocol Versioning:** Maintainers agreed on using a special version string,
|
||||
`DRAFT-2026-V1` (represented as uppercase draft, dash, 2026, dash, lowercase
|
||||
V1), to denote the new protocol version during testing.
|
||||
|
||||
## **E-Tag Correctness Mechanism & Standard IO Challenges**
|
||||
|
||||
* **Correctness vs Caching:** Geoff Goodman proposed an E-tag mechanism (or
|
||||
fencing token) where a tools list provides identifiers, and the client
|
||||
supplies that identifier on tool calls. This allows the server to verify if
|
||||
the LLM is using valid tool definitions.
|
||||
* **Tension Point \- Opaque Strings and State:** Andreas Schlapbach and Stephen
|
||||
Halter raised concerns that returning opaque strings could act as a backdoor
|
||||
to stateful behavior. However, Geoff Goodman and Stephen Halter favored
|
||||
server-defined opaque strings for simplicity.
|
||||
* **Tension Point \- Standard IO Limitations:** Standard IO lacks a native
|
||||
headers mechanism to convey E-tags, leading to a discussion on putting
|
||||
metadata in `_meta`. This prompted core maintainers to suggest investigating
|
||||
"HTTP over standard IO" to unify standard IO and HTTP features.
|
||||
|
||||
## **Future Protocol Feature Agendas**
|
||||
|
||||
* **Streaming & Binary Handling:** Peter Alexander proposed streaming results
|
||||
and raw binary handling for large payloads as priorities for future releases,
|
||||
intersecting with the file upload working group.
|
||||
* **Compression & Encryption:** Andreas Schlapbach proposed supporting message
|
||||
compression (e.g., GZip, Brotli) and message signing/encryption to guarantee
|
||||
message integrity in regulated environments.
|
||||
* **Tension Point \- Encryption Utility vs Observability:** Geoff Goodman and
|
||||
Stephen Halter questioned the utility of encryption over TLS, arguing it adds
|
||||
payload opacity and a loss of observability. Kurtis Van Gent emphasized that
|
||||
without active client demand, encryption should not be a priority.
|
||||
|
||||
# **Next Steps & Timeline**
|
||||
|
||||
Core maintainers and SDK developers will finalize conformance tests for the
|
||||
stateless protocol by the May 15th deadline. Geoff Goodman will draft an SEP
|
||||
outlining the E-tags mechanism focusing on optimistic concurrency control.
|
||||
Andreas Schlapbach will contact Saxo Bank to determine if highly regulated
|
||||
financial institutions have verified use cases for payload encryption.
|
||||
|
||||
# **Action Items**
|
||||
|
||||
- [ ] **Andreas Schlapbach** Post comments regarding the stateless SEP document,
|
||||
focusing on the initialization requirements.
|
||||
- [ ] **Stephen Halter** Implement the protocol version string `DRAFT-2026-V1`
|
||||
in the SDK and ensure conformance tests recognize it.
|
||||
- [ ] **Geoff Goodman** Draft a new SEP outlining the E-tags mechanism for
|
||||
optimistic concurrency control.
|
||||
- [ ] **Andreas Schlapbach** Talk to Saxo Bank contacts regarding encryption and
|
||||
compression messaging needs and return with example user journeys.
|
||||
|
||||
Source: [Meeting Notes — April 29,
|
||||
2026](https://docs.google.com/document/d/17snXKCMmMRtpWhRk6itCNIQ6ovTU1LmxGwumGm8FzNY/edit?usp=meet_tnfm_email)
|
||||
Loading…
Reference in a new issue