robert engels
robert engels
It looks like you would have to call a to/from method for every field access.
Which would also then require exposing read and write access to the internal long. But it would be better than encoding as string since it would be more efficient.
@k1ng440 I understand, but if you notice Timestamp is an approved 'known type' by the protobufs project and that is why it is directly (and easily) supported. protobufs would need...
closing this as infeasible
The offending code is in nghttp2_session.c:1814 ``` window_size = nghttp2_session_enforce_flow_control_limits( session, stream, NGHTTP2_DATA_PAYLOADLEN); ``` NGHTTP2_DATA_PAYLOADLEN is a constant of 16k - the framesize minimum.
I am not sure the PR is the best solution. There is a dependency on the write buffer size. If this is too small, the call fails with NGHTTP2_ERR_WOULDBLOCK and...
I don't disagree in the case of nghttp - but it's interface is not exactly "standard". And a few counter points: 1. Both Java and Go clients benefit tremendously from...
As I pointed out, the current frame size handling in nghttpd is broken. I created a PR #1650 to demonstrate. You can run this in conjunction with the -f option...
You don't need to have a larger write buffer everywhere. Why not recreate the write buffer for the connection when the server receives a max_frame_size setting update? In any case,...
Also see https://code.visualstudio.com/docs/editor/settings-sync#_sync-user-global-state-between-machines