I need to read through CEP-50 to understand the implications for that
specifically Abe but just based on our conversation so far I see no problem
with READY messages sending back a custom payload which has a list of
warnings that can be ignored by (a) earlier drivers or (b) drivers which
don't s
Maybe I'm missing something but is there really a conflict between what
you're proposing and v5 spec?
The spec seems to say two things that are immediately relevant to the
point about custom payloads in requests:
- The default QueryHandler impl just ignores them
- Only a few requests
I don’t think there’s any ambiguity in the spec for empty response types, they
should still respect header flags:
https://github.com/apache/cassandra/blob/4a548b0f289a94ce466011dba2579dcaafdba4b9/doc/native_protocol_v5.spec#L446
Is there any risk to adding a custom payload to a message type that is
explicitly documented as having an empty response body?
Thanks -- Joel.
On 8/27/2025 2:59 PM, Abe Ratnofsky wrote:
Bret - I agree with your interpretation of the spec. “Currently only X is
supported” is not a prescriptive
Bret - I agree with your interpretation of the spec. “Currently only X is
supported” is not a prescriptive statement about the protocol, as I had
initially read it.
We should encourage client implementations to log client warnings and expose
the custom payload for the READY response after proto
The v5 native protocol spec currently only supports custom payloads on query
execution messages:
https://github.com/apache/cassandra/blob/61b3d5a54befe440044ad86159a62fee487229eb/doc/native_protocol_v5.spec#L297-L298
> Currently, only QUERY, PREPARE, EXECUTE and BATCH requests support custom
