Would it be feasible to introduce either of two changes:
-
A read_timeout setting that defaults to the regular timeout but is used as an override when HTTP::Tiny::Handle is instantiated
-
The ability to specify a class to use in place of HTTP::Tiny::Handle (which would allow implementors to subclass H::T::Handle to achieve a custom tiemout)
Right now the same timeout length is always used for the connect as for reads, though H::T::Handle accepts its own timeout and thus could accommodate a distinct value.