Peter Koch writes:
So that would be a resolver that implements at least part of TCP on its own?
No, anything that uses nonblocking I/O would be vulnerable to this
programming error. It's not unusual. To name one well-known product:
the Cisco PIX had a similar error in its SMTP implementation a year or
three ago.
It could be that on some platforms, code using blocking I/O could be
vulnerable too. It's been a few years since I looked into this in
detail. Suffice to say: It's an easy mistake to make.
Arnt