code
code
> It seems like the issue here is separatation of concerns: we want to be able to make changes to the internals of the Json::Streamer without worrying about how it...
> I don't think it needs to be refactored. My proposal is to add a type in the accesslog namespace that meets the requirements of OutputBufferType in StreamerBase and to...
> If we do go with (b) I'm inclined to revert the 2 PRs that made the serializer templated on string-vs-buffer; that's complexity that would not be needed. I am...
I re-checked the **keep_value_type**'s usage again. It's `true` in most cases and will be false only when the deprecated `json_format` (was deprecated at #11125, 4 years ago) is used. So,...
I agree this is essential functionality. I can help with the review if someone want to contribute.
Sorry, I forget to submit the comments yesterday. But the only concern is the potential break change to our API and users. Could we considering add a new stats rather...
The worst case is the change may result in new version gateway/sidecar couldn't bootstrap as expected if a big hello size is configured. Although the previous config (for example with...
> With the new test that builds a 17000-byte client hello, what is the behavior without the source change? What error path does it hit? If users configured 17k max...