Amen to this. I like Qmail, but its logging sux. While the stock config
of "no to little information" is fine for a few users, whenyou start
getting into serious numbers of users, or tracking down the flow of
messages, it's non-trivial with qmail's logging...
On 24 Dec 1998, Scott Schwartz wrote:
> "D. J. Bernstein" <[EMAIL PROTECTED]> writes:
> | Scott Schwartz writes:
> | > I've already explained what I want, if you'll recall.
> |
> | Sorry, I don't recall.
>
> I don't believe you.
>
> | What exactly do you want to see in the log?
> |
> | I have asked you this question far too many times now, Scott, without
> | seeing you answer it.
>
> Didn't you see the patches I posted? In what way were they not exact
> enough?
>
> | This is your last chance.
>
> Last chance? You've already decided to ignore me. What's the point?
>
> But ok, I'll play along, just to impress you with unreciprocated
> sincerity. I'll describe the idea in english this time, with some
> improvements over the patches I previously posted. If I've made any
> typos or other obvious errors, be sure to flame them as much as
> possible, to the exclusion of the real issues at hand.
>
> If qmail-smtpd detects a timeout, it should print a message
> to stderr saying that it got a timeout, and what it was waiting
> for. (e.g. "timeout expecting CRLF.CRLF")
>
> If qmail-smtpd can't allocate memory, it should print a message
> to stderr saying that, and how many bytes it was unable to get.
> (e.g. "out of memory allocating 123456789")
>
> If it can't read a control file, it should say so, to stderr.
> (e.g. "unable to read rcpthosts")
>
> If it can't figure out its ip address, it should say so, to stderr.
> (e.g. "unable to figure out my IP addresses").
>
> If it decides to reject a message because it jas a ^J in a
> place you don't like, it should say so, to stderr.
> (e.g. "straynewline in ...")
>
> If it decides to reject a sender because it is my badmailfrom list,
> it should say so, to stderr.
> (e.g. "badmailfrom [EMAIL PROTECTED]").
>
> If it decides to reject a message because a recipient isn't in rcpthosts,
> it should say so.
> (e.g. "nogateway [EMAIL PROTECTED]").
>
> If it notices that a command is unknown, it should say so.
> (e.g. "unimplemented 'debug'").
>
> If a message has enjoyed too many hops, that fact should be logged
> to stderr.
> (e.g. "too many hops, FROM: [EMAIL PROTECTED]")
>
> If a message is too large, that fact should be noted to stderr.
> (e.g. "too big, 12345678 bytes")
>
> If it is happy with a message, the queue id should be logged
> to stderr.
>
> Naturally each message should be prefixed with a tag that identifies
> its source, such as the pid (although it would be better for tcpserver
> to setenv a session id for its children to use, so that they can fork
> without breaking the chain of custody.)
>
> If I've forgotten something, I reserve the right to revise
> and extend my remarks, and I yield back the balance of my time.
>
>
