The footnote claiming that reading the fd_set takes time proportional to the number of fds is confusing, since both the sentence referencing that footnote and the example right above that sentence are saying that the time is proportional to the largest fd. Of course it's possible to do better by storing the fds separately, but the footnote should not imply that this was done here.
(found using cat *.txt |tr -cs A-Za-z0-9_ '\n' |sort |uniq -c |sort -n |less)
The doxygen correctly described its allowed values as 1/0, but we were saying here that it took true/false.
fix type-o regarding the active port in the forking ROT13 example.
… quite often.
Fix invalid link variable to DNS document.
Several times I have seen event_active being misconstrued as a method of signaling. I've added a warning section for this function about possible resource exhaustion if not used properly.
Erroneously resetting n_written and buffer_used back to 1 would leave a gap at the beginning of the buffer, filled with the very first character that was received.