a) what should clients assume if the stream back from a HTTP GET
doesn't contain a Content-Type: header, and
b) what should a server assume, in determining what Content-Type
it should return, for files that it is serving.
These are certainly different questions: the first is important for
the design of the PROTOCOL while the second is important for the
design of a class of HTTP servers that implement the protocol.
The answer for (a) has to be the same for all servers/clients on the
network. The answer for (b) could easily vary for different
platforms, sets of documents, operating systems, etc.