On 2007-01-05 10:40:25 -0500, Guy Hulbert wrote: > On Fri, 2007-05-01 at 16:26 +0100, Peter J. Holzer wrote: > > On 2007-01-05 05:26:20 -0500, Guy Hulbert wrote: > <snip> > > > > Even if a standard is de facto instead of official, it's still a > > > > standard. /usr/sbin/sendmail certainly is the de facto standard, even > > The only argument I am making is that '/usr/sbin/sendmail' is NOT a > "standard interface" because someone was advocating that a general > solution to a problem with qpsmtpd (which I don't specifically remember) > was to modify a system by replacing it.
You may want to read the thread again. There was no problem with
qpsmtpd, therefore no solution was necessary. All well-known MTAs for
Unix-like systems come with a binary '/usr/sbin/sendmail' which
implements the same interface for sending mails as Eric Allman's
sendmail package.
> In a follow-up, I demonstrated that the only reason the Debian qmail
> package works with cron is because the /usr/lib/sendmail sym-link is
> replaced by the qmail package installation.
I don't see how you can have demonstrated that if Debian's Vixie cron
doesn't invoke /usr/lib/sendmail at all - it invokes /usr/sbin/sendmail.
There is a symlink /usr/lib/sendmail -> /usr/sbin/sendmail on Debian for
compatibility with the FHS, but cron doesn't need it.
> This is NOT the same as just dropping in the qmail
> sendmail binary into /usr/sbin/sendmail.
It is. Since /usr/lib/sendmail is simply a symlink to
/usr/sbin/sendmail, you need to replace only the latter.
> I believe that it may be a fair statement to say it is a "standard LINUX
> interface",
Nope. Standard Unix interface has been correct for at least 10, maybe 15
years. Some Unixes I've worked with in the late 1980s and early 1990s
may not have had /usr/lib/sendmail, but then they didn't have any
networking, so connecting to port 25 wouldn't have worked either (they
did have local mail, but with significant differences in
implementation).
> which is why I asked which standard was being referred to rather than
> having to check.
Not P.1003.1 - that defines only the mailx command, neither mail nor
sendmail. It's a de-facto Unix standard, not a standard defined by some
X/Open, IEEE or other recognized standardization body.
The linux FHS mandates that /usr/lib/sendmail must be a symlink to
/usr/sbin/sendmail if the latter exists and that "systems using a
sendmail-compatible mail transfer agent must provide /usr/sbin/sendmail
as a symbolic link to the appropriate executable." (but only in a
footnote).
> > > It doesn't concern me what other people do to their systems but I would
> > > like to avoid more confusion in an area which confused enough already.
> >
> > qpsmtpd was designed as a drop-in replacement for qmail-smtpd.
>
> Yes. NOT /usr/sbin/sendmail.
Right. And therefore replacing qmail-smtpd with qpsmtpd has absolutely
NO impact on programs using /usr/sbin/sendmail.
> And, BTW, there are plugins which bypass qmail-queue
I know. I wrote one of them (postfix-queue).
> which is not the same thing as qmail-smtpd.
Nobody ever claimed that.
hp
--
_ | Peter J. Holzer | I know I'd be respectful of a pirate
|_|_) | Sysadmin WSR | with an emu on his shoulder.
| | | [EMAIL PROTECTED] |
__/ | http://www.hjp.at/ | -- Sam in "Freefall"
signature.asc
Description: Digital signature
