Hey

I just like to forward the comment I entered into the bug report 519392, so
that you all are made aware of this situation. The full bug report is found
here:
http://sourceforge.net/tracker/?func=detail&atid=100976&aid=519392&group_id=976

[quote from the report follows here]

You address a valid concern here. The idea is that all callers of
Curl_write() should be modified to deal with the fact that it may return -1.
This is because I don't want anyone to loop because of this, as when we
introduce the multi interface we really want as little "looping" as possible
and if it can't be done right away we should back off and wait until we can
do it again. For the cases where this can be done of course.

The add_buffer_send() for example should be modified to instead of going into
a loop, it should store the buffer somewhere and have it re-sent later in the
main Transfer() loop as soon as it can write to the socket again.

I haven't addressed the telnet places because the telnet isn't working really
good anyway for anyone to use it and the Curl_sendf() has been an oversight.
It will need a loop in a similar manner that add_buffer_send() already has.

Thanks for pointing these out. If you have time and feel like it, I would
appreciate patches that fixes these problems.

--
 Daniel