From: Wietse Venema (wietseporcupine.org)
Date: Thu Aug 24 2006 - 12:08:02 CDT

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Jordan Russell:
> Wietse Venema wrote:
> > That would be a bug in the library. The solution is to send a bug
> > report and request that it be fixed.
>
> Supposing I did that, and both the Linux and BSD folks agreed that it
> actually was a bug, I'd also have to convince the OS vendors to backport
> the fix and put out updated libc packages for their existing releases.
> That I don't really see happening.

This happens all the time. If users don't give feedback then bugs
won't be fixed.

> If you don't believe that this should be addressed at the Postfix level
> -- even if only as a temporary workaround until the OS vendors get their
> act together -- then perhaps it would be wise to remove the claim in the
> reject_unknown_reverse_client_hostname docs that 450 is "always"
> returned for temporary problems, since what happens in the real world on
> all existing releases of Linux and BSD is just the opposite.

I could spend the rest of my life adding disclaimers with "oh by
the way don't blame Postfix if the OS produces incorrect answers"
for every feature that Postfix implements.

> In any event, I will run this by the glibc maintainers and see what they
> have to say.

A dead name server must not be a treated as if it is a permanent
error condition, expecially when claiming that a name or address
does not exist. They made the same mistake with a dead name service
for /etc/passwd, which is another cause of mail getting rejected
for the wrong reason.

        Wietse

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]