At 12:07 PM Thursday 5/6/99, [EMAIL PROTECTED] wrote:
>Mark Delany <[EMAIL PROTECTED]> writes on 5 May 1999 at 15:19:36 -0700
>
> > Perhaps it's yet again the case of an rpm-styled installed doing
> > "almost the right thing". (I confess to continually seeing reasons
> > why DjB was/is so paranoid about others packaging up qmail).
>
>Yes, BUT...*everything else* on this system (that I didn't write
>myself) was installed from RPMs, and it *all* works. I've never made
>an RPM myself, and I don't mean to cast aspersions on those who made
>the qmail RPMs. But the evidence suggests either that qmail is
>somehow harder to install right than anything else (including
>sendmail), or else that the RPMs for qmail are all badly made.
Good lord. What evidence is that exactly? Note that I only made an allusion,
not an insinuation and I specifically said "rpm-styled" to indicate that it
wasn't RPM per se. More that people who make packages need to ensure that
their startup and shutdown scripts work properly with the s/w they install.
In fact, in this case it turns out that the problem wasn't related to
packaging at all. Rather, it was the assumption that killall is universally
the right way to stop any application on Unix. As others has pointed out,
this is not the "right" way to shutdown qmail.
>If qmail is harder to install, is it for good, solid reasons, or is it
>the strain of attempting to satisfy Dan's whims? Clearly there *are*
>some good solid reasons for some of the strange things about qmail.
How you conclude that a poorly programmed package that sends the wrong
signals to the wrong processes is somehow the problem of the program that
receives the signal is well beyond me. The man page for qmail-send clearly
states what signals do what. Or is it now the case that package makers
should be excused from reading the documentation and understanding the s/w
as well?
>Since a number of different people have made qmail RPMs, it seems
>somewhat unlikely that they're *all* bad.
Who said they were? But is it possible that they are not all good?
If you've been watching this list you'll surely agree that there have been
numerous cases recently where the installation package has *not* done the
right thing and has caused more confusion that it has solved.
>Dan may believe his restrictions on distribution serve to protect
>qmail's reputation (by preventing a bad distribution from having
>security holes, for example). I'm quite sure, from the questions and
>problems I see people having, that in fact qmail's reputation is being
>seriously damaged by the lack of standard binary packages for it,
>especially a really clean RPM for the Linux world.
This has nothing whatsoever to do with anything. If you send the wrong
signal to sendmail it does not shutdown cleanly either. If you send the
wrong signal to init the system doesn't shutdown cleanly either. So it goes
for qmail. The particular distribution restrictions that Dan applies have no
bearing on this aspect whatsoever.
Regards.
