Another idea to keep in mind wrt libcurl's involvement in the user-agent
header: The header can specify the whole stack of software involved in
creating the request, so it would be nice if there were a way to tell
libcurl to add its appropriate identifier to the end of the string
provided by the libcurl user.

My libcurl-using code builds that itself today, using curl_version_info()
to get version information for the libcurl part of it.

It also builds the header itself and uses CURLOPT_HTTPHEADER, so I
hope if libcurl gets into supplying its own user-agent header, it will
recognize that.

-- 
Bryan Henderson                                   San Jose, California