Closed
Description
What is the problem this feature will solve?
This issue is similar to #42579.
We know that the header serialization will be treat differently when the header of Transfer-Encoding: chunked exists or not.
Being two different behavior for serialize the header is causing problem if people do not know how it works.
So, here is an ask for unify the behavior between two.
Related fastify/fastify-reply-from#287
cc @SimenB @mcollina @ShogunPanda
What is the feature you are proposing to solve the problem?
It should either throw with proper error or convert to latin1 in all case.
I would prefer throwing with an proper explanation.
What alternatives have you considered?
Always check in user-land code, but it is problematic and easily to forgot.