On Tue, 19 Aug 2008, Wietse Venema wrote: > Matthias Andree: > > If Postfix (or qmail[1], or whatever application) claims to support a > > particular operating system (Linux, Solaris - rather than POSIX), then > > it has to make proper assumptions to work in that possibly different > > environment that Postfix claims to support, rather than blame the OS for > > change, development, its developers for design decisions, or whatever. > > By this reasoning, I would have to evaluate kernel implementations > forever, for all the supported systems, just in case someone decides > to change the semantics of a system call and while still claiming > that link(2) is POSIX conformant.
True. Or declare Postfix supported only on systems fully conforming to POSIX - which is not the route you chose to go (luckily, because that's also impractical). > Likewise, I would have to evaluate ANSI C environments and ensure > that some implementation does not introduce a change. > > That is obviously not practical. So it will obviously not happen, and I appreciate that because it leaves free Wietse brain cycles for Postfix rather than his wasting energy on a regular OS review. My point being that someone will have to take responsibility to fix the particular incompatibility between application expectations and OS provisions. You fixed your applications. Many thanks for that, and thanks for not passing the buck and leaving the issue unfixed. Other authors decided to pass the buck, their applications remaining unfixed after a dozen years... that is also not practical. Thanks again. -- Matthias Andree
