On 25 Jan 2001, D. J. Bernstein wrote:
> In fact, it's not a bug; it's a portability problem. If you were using
> OpenBSD, you'd see outgoing connections to 0.0.0.0 rejected with EINVAL.
This OpenBSD idiosyncracy is almost exactly two years old [1], i.e.
OpenBSD 2.4 and earlier are affected (well, sane people have probably
upgraded in the meantime). It isn't even documented properly, their
connect(2) [2] says:
[EINVAL] A TCP connection with a local broadcast, the all-ones or a
multicast address as the peer was attempted.
In fact, they did not even bother to mention the change in their
Daily Changelog [3] and CVS log entry say "netinet merge of NRL
stuff. some indent and shrinkage needed; NRL/cmetz". And the funny thing
is that everyone appears to call the equivalence of 0.0.0.0 and 127.0.0.1
for TCP connects a *BSDism* (undocumented, as usual), ergo the change
does probably qualify as "a frivolous incompatibility."
Now, how old qmail 1.03 is? CHANGES in qmail-1.03.tar.gz say it was
released on June 15 1998. Hmm...this predates the change in question
(January 11 1999), doesn't it? Did you code qmail with a crystal ball in
your hand? With all due respect, aren't you just looking for lame excuses
(like playing with words and renaming bugs to portability problems) in
order not to have to admit there is even the slightest imperfection in
your creation?
[1] http://www.openbsd.org/cgi-bin/cvsweb/src/sys/netinet/tcp_usrreq.c?r1=1.31&r2=1.32
[2]
http://www.openbsd.org/cgi-bin/man.cgi?query=connect&apropos=0&sektion=0&manpath=OpenBSD+Current&arch=i386&format=html
[3] http://www.openbsd.org/plus25.html
--Pavel Kankovsky aka Peak [ Boycott Microsoft--http://www.vcnet.com/bms ]
"Resistance is futile. Open your source code and prepare for assimilation."
