Malte,
good catch. I have actually no idea, why "accept: */*" was there (the rfc says, it is optional). It is certainly not a good idea to hard-code the transmission.
Since i see no good reason to set this header-field differently than other request header fields, i took another approach and removed the automatic transmission of "accept" altogether. If a client wants to specify "accept", it can be done - like for all other header fields - via request_header_fields, seems that it was your first guess as well.
-gustaf neumann
PS: fixed in CVS head.