CURLOPT_DEBUGFUNCTION and CURLOPT_DEBUGDATA.

do if the write would've blocked

on my test machine! :-)

This was a real bummer!

boxes.

cases where EWOULDBLOCK or equivalent is returned. We must not block.

over to non-blocking sockets, but this loops very nastily. We should return
back to the select() and wait there until more data arrives, not just blindly
attempt again and again...

for each *_perform(). It makes things a lot easier, as the first one that
detects the error get to write the final error reason...

be posted in a minute to the libcurl list.

  set CURLOPT_WRITEHEADER to non-NULL, even if you didn't care about that
  data. This is now fixed.

"low leve" interface is dumped

for FTP-stuff

Curl_write() and Curl_read() are here

firewalls (like FW-1) seems to dislike split-up writes at times...